JP7046007B2 - How to adjust the molecular label count - Google Patents

Description

Related application This application is based on 35 U.S. S. C. According to §119 (e), US Provisional Patent Application No. 62/342137 filed May 26, 2016; US Provisional Patent Application No. 62/381945 filed August 31, 2016; and 2016. Claim priority under US Provisional Patent Application No. 62/401720, filed September 29. The contents of each of these applications are expressly incorporated by reference in their entirety with this application.

The present disclosure relates generally to the field of nucleic acid barcoding, more specifically PCR and sequencing error correction using molecular labeling.

Description of Related Fields Methods and techniques such as probabilistic barcoding use in cell analysis, for example, reverse transcription, polymerase chain reaction (PCR) amplification, and next-generation sequencing (NGS) to determine cell status. Therefore, it is useful in decoding the gene expression profile. However, these methods and techniques can introduce errors such as substitution errors (including one or more bases) and unsubstituted errors, which, if left uncorrected, can result in overrated molecular counts. .. Therefore, there is a need for methods and techniques that can correct various errors in order to obtain accurate molecular counts estimated using probabilistic barcoding.

A method of determining the number of targets is disclosed herein. In some embodiments, the method is (a) a step of using a plurality of probability barcodes to attach probability barcodes to a plurality of targets to generate a plurality of targets with probability barcodes. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic bar code; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; and a step of determining the quality status of the target in the sequencing data obtained in (ii) (b); (iii). ) The step of determining one or more sequencing data errors in the sequencing data obtained in (b), wherein the step of determining one or more sequencing data errors in the sequencing data is as follows: The number of molecular labels with identifiable sequences associated with the target in the sequencing data, the quality status of the target in the sequencing data, and the number of molecular labels with identifiable sequences in multiple probability barcodes. A step comprising determining one or more of them; (iv) a step of estimating the number of targets, wherein the estimated number of targets is one or more sequencing data errors determined in (iii). Includes (i) a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted according to. Steps (i), (ii), (iii), and (iv) can be performed for each of the plurality of targets. The method can be multiplexed.

In some embodiments, the method further comprises collapsing the sequencing data obtained in (b) before determining one or more sequencing data errors. In the step of collapsing the sequencing data obtained in (b), a copy of a target having a similar molecular label and having an occurrence number smaller than a predetermined folding occurrence number threshold is printed on the same molecular label for a plurality of targets. Including the step of assigning as having, where the two copies of the target have similar molecular labels if the sequences of the molecular labels of the two copies of the target differ by at least one base.

In some embodiments, the predetermined folding occurrence threshold can be 7 if the probability barcode contains about 6651 molecular labels with identifiable sequences. If the probability barcode contains about 65536 molecular labels with identifiable sequences, the predetermined folding occurrence number threshold can be 17. The two copies of the target have similar molecular labels if the sequences of the molecular labels on the two copies of the target differ by at least one base. In some embodiments, the molecular label comprises 5-20 nucleotides. The molecular labels of the various probability barcodes may be different from each other. Multiple probability barcodes include about 6651 molecular labels with identifiable sequences. Multiple probability barcodes include about 65536 molecular labels with identifiable sequences.

In some embodiments, it comprises a sequence of a plurality of targets having a read length of 50 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 75 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 100 nucleotides or greater. The sequencing data obtained in (b) can be generated by performing polymerase chain reaction (PCR) amplification on a plurality of probabilistic barcoded targets.

In some embodiments, the one or more sequencing data errors can be PCR introduction errors, sequencing introduction errors, errors due to bar code contamination, library fabrication errors, or any combination thereof. PCR induction errors can be the result of PCR amplification errors, PCR amplification bias, inadequate PCR amplification, or any combination thereof. Sequencing implementation errors can be the result of inaccurate base calling, inadequate sequencing, or any combination thereof.

In some embodiments, the quality status of the target in the sequencing data can be complete sequencing, incomplete sequencing, or saturated sequencing. The quality status of a target in the sequencing data is the number of molecular labels that have an identifiable sequence in multiple probability barcodes and the molecule that has an identifiable sequence associated with the target in the counted sequencing data. It can be determined by the number of signs. The quality status of the target in the sequencing data is defined as incomplete sequencing when the quality status of the target in the sequencing data obtained in (b) is not complete sequencing and is not saturated sequencing. Can be categorized.

In some embodiments, the complete sequencing quality status is determined by a dispersal index compared to a Poisson distribution above a predetermined complete sequencing application threshold, where the predetermined complete sequencing application threshold is 0.9, It can be 1 or 4. The complete sequencing quality status can also be further determined by a molecular label having an incidence greater than or equal to a predetermined complete sequencing occurrence number threshold in the sequencing data obtained in (b), where the predetermined completeness. The sequencing occurrence number threshold can be 10 or 18.

In some embodiments, the saturation sequencing quality status can be determined by a target having a number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. Saturation sequencing quality status can also be further determined by one of the other targets having a number of molecular labels containing identifiable sequences greater than a predetermined saturation threshold. The predetermined saturation threshold can be 6557 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The predetermined saturation threshold can be 65532 if the probability barcode contains about 65536 molecular labels with identifiable sequences.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is, in (iv), that the target has a complete sequencing quality status. If so, a step of determining all child molecule labels for one or more parent molecule labels; and a step of performing a first statistical analysis on at least one child molecule label and parent molecule label; a first statistic. If the null hypothesis of the analysis is accepted, it is regulated by the step of assigning the number of child molecule labels to the parent molecule label.

In some embodiments, one or more parent molecular labels include a molecular label having an incidence greater than or equal to a predetermined complete sequencing parent threshold, wherein the predetermined complete sequencing parent threshold is a predetermined complete sequence. Equal to the number of sing occurrence thresholds. The child molecule label comprises a molecule label that is one base different from the parent molecule label and has a number of occurrences that is less than or equal to the predetermined complete sequencing child threshold, where the predetermined complete sequencing child threshold is 3 or 5. It is possible. The null hypothesis of the first statistical analysis is acceptable if the probability of the null hypothesis that it is true is less than the false discovery rate, where the false discovery rate is 5% or 10%. The first statistical analysis may be a multiple binomial test.

In some embodiments, the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), in (iv), the target has a complete sequencing quality status. If so, it is regulated by the step of thresholding the molecular label of the target to determine the true and false molecular label associated with the target in the sequencing data obtained in (b). The step of thresholding the target molecular label comprises performing a second statistical analysis on the target molecular label.

In some embodiments, the steps of performing a second statistical analysis are as follows: fitting the distribution of target molecular labels and the number of their occurrences into two Poisson distributions; true using two Poisson distributions. It comprises a step of determining the number n of molecular labels; a step of removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label is the nth richest molecule. A true molecular label comprises a molecular label having a number of occurrences lower than the number of occurrences of the label, and a true molecular label includes a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label. The two Poisson distributions include a first Poisson distribution corresponding to a true molecular label and a second Poisson distribution corresponding to a false molecular label.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is, in (iv), the number of sequencing obtained in (b). If the quality status of the target in the data is incomplete sequencing quality status, the step of determining whether the target is noisy in the sequencing data obtained in (b); It can be regulated by the step of removing the noisy target from the sequenced data obtained. If the number of molecular labels generated by the noisy target is less than or equal to the incomplete sequencing noisy target threshold, the target may be noisy, where the incomplete sequencing noisy gene threshold is 5. The incomplete sequencing noisy target threshold may be equal to the median or average incidence of molecular labels for multiple targets with complete sequencing quality status.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is in (iv) the sequencing obtained in (b). If the quality status of the target in the data is incomplete sequencing quality status, the molecular label of the target is thresholded to determine the true and false molecular labels in the sequencing data obtained in (b). It can be adjusted by the process of

In some embodiments, the step of thresholding the target molecular label comprises performing a third statistical analysis on the molecular label. The step of performing the third statistical analysis on the molecular label is the step of determining the number n of true molecular labels using the zero-cleaving Poisson model; and the false molecular label from the sequencing data obtained in (b). The false molecular label comprises a molecular label having a lower number of occurrences than the nth abundant molecular label, and the true molecular label comprises the nth. Includes molecular labels with more than abundant molecular labels.

In some embodiments, the sequencing data counted in (i) is adjusted according to one or more sequencing data errors determined in (iii) and then the sequence obtained in (b). At least 50% or 80% of the molecular labels in the singing data can be retained.

In some embodiments, the step of attaching a probability barcode to a plurality of targets comprises hybridizing the plurality of probability barcodes with the plurality of targets to generate a target with a probability barcode. The step of attaching a probability barcode to a plurality of targets includes a step of creating an indexed library of targets with a probability barcode. The step of creating an indexed library of targets with probability barcodes can be performed using a solid support containing a plurality of probability barcodes. The solid support comprises a plurality of synthetic particles associated with a plurality of probability barcodes. The solid support comprises multiple two-dimensional or three-dimensional probability barcodes. Solid carriers include polymers, matrices, hydrogels, needle array devices, antibodies, or any combination thereof.

In some embodiments, each of the plurality of probability barcodes comprises one or more of a sample label, a universal label and a cell label, wherein the sample label is for the plurality of probability barcodes on a solid carrier. It may be the same, the universal label may be the same for multiple probability barcodes on a solid carrier, and the cell label may be the same for multiple probability barcodes on a solid carrier. The sample label contains 5-20 nucleotides. The universal label contains 5-20 nucleotides. Cell labels contain 5-20 nucleotides.

In some embodiments, the synthetic particles may be beads. The beads may be silica gel beads, porous glass beads, magnetic beads, dyna beads, sephadex / sephadex beads, cellulose beads, polystyrene beads, or any combination thereof.

In some embodiments, multiple targets can be included in the sample. The sample contains one or more cells. The sample may be a single cell. One or more cells include one or more cell types. At least one of one or more cell types can be brain cells, heart cells, cancer cells, circulating tumor cells, organ cells, epithelial cells, metastatic cells, benign cells, primary cells, circulating cells, or any combination thereof. be.

In some embodiments, the plurality of targets are ribonucleic acid (RNA), messenger RNA (mRNA), microRNA, small interfering RNA (siRNA), RNA degradation products, RNA containing poly (A) tail, or RNA thereof. Includes any combination of.

In some embodiments, the method may further comprise the step of lysing one or more cells. The steps of lysing one or more cells include heating the sample, contacting the sample with a detergent, changing the pH of the sample, or any combination thereof.

A method of determining the number of targets is disclosed herein. In some embodiments, the method is (a) a step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a plurality of targets with the probability barcodes. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic barcode; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; (iii) a step of identifying clusters of target molecular labels using directional proximity; (iii) (ii). ), Using the cluster of molecular labels of the targets identified in), the step of collapsing the sequencing data obtained in (b); (iv) the step of estimating the number of targets, the estimated number of targets. Includes (ii) folding of the sequencing data followed by a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i). Multiple targets include targets for the entire transcriptome of the cell.

In some embodiments, the target molecular labels within the cluster are within a predetermined directional proximity threshold of each other. The directional proximity threshold is a Hamming distance of 1. The target molecular label within the cluster comprises one or more parent molecule labels and a child molecule label of one or more parent molecule labels, wherein the number of occurrences of the parent molecule label is the number of occurrences of predetermined directional proximity. It is above the threshold. The predetermined number of directional proximity occurrence thresholds may be 2 × ( number of occurrences of child molecule labels) -1 .

In some embodiments, the step of collapsing the sequencing data obtained in (b) using the cluster of target molecular labels identified in (ii) converts the number of child molecule labels generated into the parent molecule label. Includes the process of attribution.

In some embodiments, the method may further include determining the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be 15-20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to carry out the statistical analysis are as follows: the steps of applying the distribution of target molecular labels and the number of their occurrences to the two Poisson distributions; and the step of determining the number n of true molecular labels using the two Poisson distributions. Includes a step of removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label has a lower number of occurrences than the nth most abundant molecular label. A true molecular label comprises a molecular label having, and a true molecular label comprises a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label.

The present specification discloses a computer system for determining the number of targets. In some embodiments, the computer system comprises one or more computer processors in contact with computer-readable memory; where the computer-readable memory stores executable instructions; where one or more computer processors execute. Programmed by possible instructions, (a) a process of attaching probability barcodes to multiple targets using multiple probability barcodes to generate multiple probability barcoded targets, with multiple probability barcodes. Each of the steps comprises a molecular label; (b) a step of obtaining sequencing data for a target with a probability bar code; (c) for one or more of multiple targets: (i) to a target in the sequencing data. A step of counting the number of molecular labels having an associated identifiable sequence; and a step of determining the quality status of the target in the sequencing data obtained in (ii) (b); (iii) (b). The step of determining one or more sequencing data errors in the sequencing data obtained in the above, and the step of determining one or more sequencing errors in the sequencing data, is as follows: One or more of the number of molecular labels with identifiable sequences associated with the target, the quality status of the target in the sequencing data, and the number of molecular labels with identifiable sequences in multiple probability barcodes. A step comprising determining; (iv) a step of estimating the number of targets, wherein the estimated number of targets is adjusted according to one or more sequencing data errors determined in (iii). It also comprises a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i). Steps (i), (ii), (iii), and (iv) can be performed for each of the plurality of targets. Steps (a), (b), (c), (i), (ii), (iii), and (iv) can be multiplexed.

In some embodiments, the executable instruction further performs one or more steps of collapsing the sequencing data obtained in (b) before determining one or more sequencing data errors. You can also program your computer processor. In the step of collapsing the sequencing data obtained in (b), a copy of a target having a similar molecular label and an occurrence number smaller than a predetermined folding occurrence number threshold is obtained, and the same molecular label is applied to a plurality of targets. Including the step of assigning as having, where the two copies of the target have similar molecular labels if the sequences of the molecular labels on the two copies of the target differ by at least one base.

In some embodiments, the predetermined folding occurrence threshold can be 7 if the probability barcode contains about 6651 molecular labels with identifiable sequences. If the probability barcode contains about 65536 molecular labels with identifiable sequences, the predetermined folding occurrence number threshold can be 17. The two copies of the target have similar molecular labels if the sequences of the molecular labels on the two copies of the target differ by at least one base. In some embodiments, the molecular label comprises 5-20 nucleotides. The molecular labels of the various probability barcodes may be different from each other. Multiple probability barcodes include about 6651 molecular labels with identifiable sequences. Multiple probability barcodes include about 65536 molecular labels with identifiable sequences.

In some embodiments, it comprises a sequence of a plurality of targets having a read length of 50 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 75 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 100 nucleotides or greater. The sequencing data obtained in (b) can be generated by performing polymerase chain reaction (PCR) amplification on a plurality of probabilistic barcoded targets.

In some embodiments, the one or more sequencing data errors can be PCR introduction errors, sequencing introduction errors, errors due to bar code contamination, library fabrication errors, or any combination thereof. PCR induction errors can be the result of PCR amplification errors, PCR amplification bias, inadequate PCR amplification, or any combination thereof. Sequencing implementation errors can be the result of inaccurate base calling, inadequate sequencing, or any combination thereof.

In some embodiments, the executable instruction further determines that the quality status of the target in the sequencing data is complete sequencing, incomplete sequencing, or saturated sequencing. You can also program your computer processor. The quality status of a target in the sequencing data is the number of molecular labels that have an identifiable sequence in multiple probability barcodes and the molecule that has an identifiable sequence associated with the target in the counted sequencing data. It can be determined by the number of signs. The quality status of the target in the sequencing data shall be classified as incomplete sequencing if the quality status of the target in the sequencing data obtained in (b) is neither complete sequencing nor saturated sequencing. Can be done.

In some embodiments, the executable instruction further comprises one or more computer processors such that the complete sequencing quality status is determined by a dispersal index compared to a Poisson distribution above a predetermined complete sequencing dispersal threshold. It can also be programmed, where the predetermined complete sequencing application threshold can be 0.9, 1, or 4. The complete sequencing quality status can also be further determined by a molecular label having an incidence greater than or equal to a predetermined complete sequencing occurrence number threshold in the sequencing data obtained in (b), where the predetermined completeness. The sequencing occurrence number threshold can be 10 or 18.

In some embodiments, the executable instruction further determines the saturation sequencing quality status by a target having a specific number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. It is also possible to program one or more computer processors. Saturation sequencing quality status can also be further determined by one of the other targets having a particular number of molecular labels containing an identifiable sequence greater than a predetermined saturation threshold. The predetermined saturation threshold can be 6557 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The predetermined saturation threshold can be 65532 if the probability barcode contains about 65536 molecular labels with identifiable sequences.

In some embodiments, the executable instruction further determines, in (iv), the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i). With the step of determining all child molecule labels for one or more parent molecule labels if they have full sequencing quality status; perform a first statistical analysis for at least one child molecule label and parent molecule label. Steps; If the null hypothesis of the first statistical analysis is accepted, program one or more computer processors to regulate by the step of assigning the number of child molecule labels to the parent molecule label. You can also.

In some embodiments, one or more parent molecular labels include a molecular label having an incidence greater than or equal to a predetermined complete sequencing parent threshold, wherein the predetermined complete sequencing parent threshold is a predetermined complete sequence. Equal to the number of sing occurrence thresholds. The child molecule label comprises a molecule label that is one base different from the parent molecule label and has a number of occurrences that is less than or equal to the predetermined complete sequencing child threshold, where the predetermined complete sequencing child threshold is 3 or 5. It is possible. The null hypothesis of the first statistical analysis is acceptable if the probability of the null hypothesis that it is true is less than the false discovery rate, where the false discovery rate is 5% or 10%. The first statistical analysis may be a multiple binomial test.

In some embodiments, the executable instruction further determines, in (iv), the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i). If it has a complete sequencing quality status, it thresholds the target's molecular label to determine the true and false molecular label associated with the target in the sequencing data obtained in (b). One or more computer processors can also be programmed to adjust according to the process. The step of thresholding the target molecular label comprises performing a second statistical analysis on the target molecular label.

In some embodiments, the viable instruction further applies the distribution of target molecular labels and the number of their occurrences to the two Poisson distributions; the number n of true molecular labels using the two Poisson distributions. One or more computer processors to carry out the step of performing the second statistical analysis by the step of determining; and the step of removing the false molecular label from the sequencing data obtained in (b). It can also be programmed, where the false molecular label contains a molecular label with a lower number of occurrences than the nth richest molecular label, and the true molecular label is the nth richest. Includes molecular labels with more than the number of molecular labels generated . The two Poisson distributions include a first Poisson distribution corresponding to a true molecular label and a second Poisson distribution corresponding to a false molecular label.

In some embodiments, the executable instruction further, in (iv), the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), (b). If the quality status of the target in the sequencing data obtained in) is incomplete sequencing quality status, the step of determining whether the target is noisy in the sequencing data obtained in (b). And; one or more computer processors can also be programmed to be tuned by the step of removing noisy targets from the sequencing data obtained in (b). If the number of molecular labels generated by the noisy target is less than or equal to the incomplete sequencing noisy target threshold, the target may be noisy, where the incomplete sequencing noisy gene threshold is 5. The incomplete sequencing noisy target threshold may be equal to the median or average incidence of molecular labels for multiple targets with complete sequencing quality status.

In some embodiments, the executable instruction further, in (iv), the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), (b). If the quality status of the target in the sequencing data obtained in) is incomplete sequencing quality stator, the molecular label of the target is thresholded and the true molecular label in the sequencing data obtained in (b). And one or more computer processors can be programmed to adjust by the step of determining false molecular labels.

In some embodiments, the executable instruction may further program one or more computer processors to threshold the target molecular label by performing a third statistical analysis on the molecular label. can. The step of performing the third statistical analysis on the molecular label is the step of determining the number n of true molecular labels using the zero-cleaving Poisson model; and the false molecular label from the sequencing data obtained in (b). The false molecular label comprises a molecular label having a lower number of occurrences than the nth abundant molecular label, and the true molecular label comprises the nth. Includes molecular labels with more than abundant molecular labels.

In some embodiments, the sequencing data obtained in (i) is adjusted according to one or more sequencing data errors determined in (iii), and then the sequencing obtained in (b). At least 50% or 80% of the molecular labels in the data can be retained.

The present specification discloses a computer system for determining the number of targets. In some embodiments, the computer system comprises a computer-readable memory for storing executable instructions; one or more computer processors in contact with the computer-readable memory, wherein the one or more computer processors are: (A) A step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a target with a plurality of probability barcodes, and each of the plurality of probability barcodes includes a molecular label. Steps and; (b) Obtaining sequencing data for targets with probability barcodes; (c) For one or more of multiple targets: (i) Identifiable sequences associated with the targets in the sequencing data. A step of counting the number of molecular labels having; (ii) a step of identifying a cluster of target molecular labels using directional proximity; and a step of identifying a cluster of target molecular labels identified in (iii) (ii). In the step of collapsing the sequencing data obtained in (b) and (iv) the step of estimating the number of targets, the estimated number of targets is the step of collapsing the sequencing data in (ii). Then, a step of correlating with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i) is programmed by a viable instruction. Multiple targets include targets for the entire transcriptome of the cell.

In some embodiments, the target molecular labels within the cluster are within a predetermined directional proximity threshold of each other. The directional proximity threshold is a Hamming distance of 1. The target molecular label within the cluster comprises one or more parent molecule labels and a child molecule label of one or more parent molecule labels, wherein the number of occurrences of the parent molecule label is the number of occurrences of predetermined directional proximity. It is above the threshold. The predetermined number of directional proximity occurrence thresholds may be 2 × ( number of occurrences of child molecule labels) -1 .

In some embodiments, the step of collapsing the sequencing data obtained in (b) using the cluster of target molecular labels identified in (ii) converts the number of child molecule labels generated into the parent molecule label. Includes the process of attribution.

In some embodiments, the executable instruction can further program one or more computer processors to determine the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be 15-20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to carry out the statistical analysis are as follows: the steps of applying the distribution of target molecular labels and the number of their occurrences to the two Poisson distributions; and the step of determining the number n of true molecular labels using the two Poisson distributions. Includes a step of removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label has a lower number of occurrences than the nth most abundant molecular label. A true molecular label comprises a molecular label having, and a true molecular label comprises a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label.

The present specification discloses one or more non-transient computer reading media containing executable code, which, when executed, causes one or more computer devices to determine the number of targets. In some embodiments, the executable code, when executed, attaches probabilistic barcodes to one or more computer devices, the following: (a) multiple probabilistic barcodes and multiple targets. A step of generating a plurality of probabilistic bar coded targets, wherein each of the plurality of probabilistic bar codes includes a molecular label; (b) a step of acquiring sequencing data of the probabilistic bar coded target; (c). ) For one or more of a plurality of targets: (i) counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data; (ii) the sequence obtained in (b). The step of determining the quality status of the target in the single data; the step of determining one or more sequencing data errors in the sequencing data obtained in (iii) (b), in the sequencing data. The steps to determine one or more sequencing errors are as follows: the number of molecular labels with identifiable sequences associated with the target in the sequencing data, the quality status of the target in the sequencing data, and multiple probabilities. A step comprising determining one or more of the number of molecular labels having an identifiable sequence on the bar code; (iv) a step of estimating the number of targets, wherein the estimated number of targets is (iv). Correlates with the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i), adjusted for one or more sequencing data errors determined in iii). And the process including. Steps (i), (ii), (iii), and (iv) can be performed for each of the plurality of targets. The method can be multiplexed.

In some embodiments, the process further comprises collapsing the sequencing data obtained in (b) before determining one or more sequencing data errors. In the step of collapsing the sequencing data obtained in (b), a copy of a target having a similar molecular label and an occurrence number smaller than a predetermined folding occurrence number threshold is obtained, and the same molecular label is applied to a plurality of targets. Including the step of assigning as having, where the two copies of the target have similar molecular labels if the sequences of the molecular labels on the two copies of the target differ by at least one base.

In some embodiments, the predetermined folding occurrence threshold can be 7 if the probability barcode contains about 6651 molecular labels with identifiable sequences. If the probability barcode contains about 65536 molecular labels with identifiable sequences, the predetermined folding occurrence number threshold can be 17. The two copies of the target have similar molecular labels if the sequences of the molecular labels on the two copies of the target differ by at least one base. In some embodiments, the molecular label comprises 5-20 nucleotides. The molecular labels of the various probability barcodes may be different from each other. Multiple probability barcodes include about 6651 molecular labels with identifiable sequences. Multiple probability barcodes include about 65536 molecular labels with identifiable sequences.

In some embodiments, the sequencing data comprises a sequence of multiple targets with a read length of 50 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 75 nucleotides or greater. Sequencing data include sequences of multiple targets with read lengths of 100 nucleotides or greater. The sequencing data obtained in (b) can be generated by performing polymerase chain reaction (PCR) amplification on a plurality of probabilistic barcoded targets.

In some embodiments, the one or more sequencing data errors can be PCR introduction errors, sequencing introduction errors, errors due to bar code contamination, library fabrication errors, or any combination thereof. PCR induction errors can be the result of PCR amplification errors, PCR amplification bias, inadequate PCR amplification, or any combination thereof. Sequencing implementation errors can be the result of inaccurate base calling, inadequate sequencing, or any combination thereof.

In some embodiments, the quality status of the target in the sequencing data can be complete sequencing, incomplete sequencing, or saturated sequencing. The quality status of a target in the sequencing data is the number of molecular labels that have an identifiable sequence in multiple probability barcodes and the molecule that has an identifiable sequence associated with the target in the counted sequencing data. It can be determined by the number of signs. The quality status of the target in the sequencing data is defined as incomplete sequencing when the quality status of the target in the sequencing data obtained in (b) is not complete sequencing and is not saturated sequencing. Can be categorized.

In some embodiments, the complete sequencing quality status is determined by the dispersion index for a Poisson distribution above a predetermined complete sequencing application threshold, where the predetermined complete sequencing application threshold is 0.9, 1, ,. Or it can be 4. The complete sequencing quality status can also be further determined by a molecular label having an incidence greater than or equal to a predetermined complete sequencing occurrence number threshold in the sequencing data obtained in (b), where the predetermined completeness. The sequencing occurrence number threshold can be 10 or 18.

In some embodiments, the saturation sequencing quality status can be determined by a target having a number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. Saturation sequencing quality status can also be further determined by one of the other targets having a number of molecular labels containing identifiable sequences greater than a predetermined saturation threshold. The predetermined saturation threshold can be 6557 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The predetermined saturation threshold can be 65532 if the probability barcode contains about 65536 molecular labels with identifiable sequences.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is, in (iv), that the target has a complete sequencing quality status. If so, a step of determining all child molecule labels for one or more parent molecule labels; and a step of performing a first statistical analysis on at least one child molecule label and parent molecule label; a first statistic. If the null hypothesis of the analysis is accepted, it is regulated by the step of assigning the number of child molecule labels to the parent molecule label.

In some embodiments, one or more parent molecular labels include a molecular label having an incidence greater than or equal to a predetermined complete sequencing parent threshold, wherein the predetermined complete sequencing parent threshold is a predetermined complete sequence. Equal to the number of sing occurrence thresholds. The child molecule label comprises a molecule label that is one base different from the parent molecule label and has a number of occurrences that is less than or equal to the predetermined complete sequencing child threshold, where the predetermined complete sequencing child threshold is 3 or 5. It is possible. The null hypothesis of the first statistical analysis is acceptable if the probability of the null hypothesis that it is true is less than the false discovery rate, where the false discovery rate is 5% or 10%. The first statistical analysis may be a multiple binomial test.

In some embodiments, the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), in (iv), the target has a complete sequencing quality status. If so, it is regulated by the step of thresholding the molecular label of the target to determine the true and false molecular label associated with the target in the sequencing data obtained in (b). The step of thresholding the target molecular label comprises performing a second statistical analysis on the target molecular label.

In some embodiments, the step of performing the second statistical analysis is to fit the distribution of the target molecular labels and the number of their occurrences to the two Poisson distributions; the true molecular label using the two Poisson distributions. A step of determining the number n of the above; a step of removing the fake molecular label from the sequencing data obtained in (b), wherein the fake molecular label is the nth richest molecular label. A true molecular label comprises a molecular label having a number of occurrences lower than the number of occurrences, and a true molecular label comprises a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label. The two Poisson distributions include a first Poisson distribution corresponding to a true molecular label and a second Poisson distribution corresponding to a false molecular label.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is, in (iv), the number of sequencing obtained in (b). If the quality status of the target in the data is incomplete sequencing quality status, the step of determining whether the target is noisy in the sequencing data obtained in (b); It can be adjusted by the step of removing the noisy target from the sequenced data obtained. If the number of molecular labels generated by the noisy target is less than or equal to the incomplete sequencing noisy target threshold, the target may be noisy, where the incomplete sequencing noisy gene threshold is 5. The incomplete sequencing noisy target threshold may be equal to the median or average incidence of molecular labels for multiple targets with complete sequencing quality status.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is in (iv) the sequencing obtained in (b). If the quality status of the target in the data is an incomplete sequencing quality stator, the molecular label of the target is thresholded to determine the true and false molecular labels in the sequencing data obtained in (b). It can be adjusted by the process of

In some embodiments, the step of thresholding the target molecular label comprises performing a third statistical analysis on the molecular label. The step of performing the third statistical analysis on the molecular label is the step of determining the number n of true molecular labels using the zero-cleaving Poisson model; and the false molecular label from the sequencing data obtained in (b). The false molecular label comprises a molecular label having a lower number of occurrences than the nth abundant molecular label, and the true molecular label comprises the nth. Includes molecular labels with more than abundant molecular labels.

In some embodiments, the sequencing data counted in (i) is adjusted according to one or more sequencing data errors determined in (iii) and then the sequence obtained in (b). At least 50% or 80% of the molecular labels in the singing data can be retained.

The present specification discloses one or more non-transient computer reading media containing executable code, which, when executed, causes one or more computer devices to determine the number of targets. In some embodiments, the executable code, when executed, attaches probabilistic barcodes to one or more computer devices, the following: (a) with multiple probabilistic barcodes and multiple targets. A step of generating a target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes includes a molecular label; (b) a step of acquiring sequencing data of the target with a probability barcode; (c). ) For one or more of multiple targets: (i) counting the number of molecular labels with identifiable sequences associated with the target in the sequencing data; (ii) using directional proximity to the target. (Iii) A step of collapsing the sequencing data obtained in (b) using the target molecularly labeled clusters identified in (iii); (iv) Target. In the step of estimating the number of targets, the estimated number of targets is identifiable associated with the targets in the sequencing data counted in (i) after collapsing the sequencing data in (ii). Perform a process comprising a step that correlates with the number of molecular labels having a sequence. Multiple targets include targets for the entire transcriptome of the cell.

In some embodiments, the target molecular labels within the cluster are within a predetermined directional proximity threshold of each other. The directional proximity threshold is a Hamming distance of 1. The target molecular label within the cluster comprises one or more parent molecule labels and a child molecule label of one or more parent molecule labels, wherein the number of occurrences of the parent molecule label is the number of occurrences of predetermined directional proximity. It is above the threshold. The predetermined number of directional proximity occurrence thresholds may be 2 × ( number of occurrences of child molecule labels) -1 .

In some embodiments, the step of collapsing the sequencing data obtained in (b) using the cluster of target molecular labels identified in (ii) converts the number of child molecule labels generated into the parent molecule label. Includes the process of attribution.

In some embodiments, the method may further include determining the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be 15-20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to carry out the statistical analysis are as follows: the steps of applying the distribution of target molecular labels and the number of their occurrences to the two Poisson distributions; and the step of determining the number n of true molecular labels using the two Poisson distributions. Includes a step of removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label has a lower number of occurrences than the nth most abundant molecular label. A true molecular label comprises a molecular label having, and a true molecular label comprises a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label.

The present specification discloses methods for correcting PCR or sequencing errors. In some embodiments, the method comprises (a) acquiring sequencing data for a probabilistic bar coded target; (b) for one or more of a plurality of targets: (i) targets in the sequencing data. Counting the number of molecular labels with identifiable sequences associated with; (ii) identifying clusters of target molecular labels using directional proximity; (iii) (ii). Using the clusters of molecular labels of the targeted targets, a step of collapsing the sequencing data obtained in (a); (iv) a step of estimating the number of targets can be included, which is estimated here. The number of targets correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i) after folding the sequencing data in (ii). Multiple targets include targets for the entire transcriptome of the cell. In some embodiments, the method can be used to determine the number of targets. The method further comprises (c) using multiple probability barcodes to attach probability barcodes to multiple targets to generate multiple probability barcoded targets; (d) probability barcoded targets. Can include a step of sequencing to generate sequencing data for a target with a received probability barcode.

In some embodiments, the target molecular labels within the cluster are within a predetermined directional proximity threshold of each other. The directional proximity threshold is a Hamming distance of 1. The target molecular label within the cluster comprises one or more parent molecule labels and a child molecule label of one or more parent molecule labels, wherein the number of occurrences of the parent molecule label is the number of occurrences of predetermined directional proximity. It is above the threshold. The predetermined number of directional proximity occurrence thresholds may be 2 × ( number of occurrences of child molecule labels) -1 .

In some embodiments, the step of collapsing the sequencing data obtained in (b) using the cluster of target molecular labels identified in (ii) converts the number of child molecule labels generated into the parent molecule label. Includes the process of attribution.

In some embodiments, the method further comprises the step of determining the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be 15-20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to carry out the statistical analysis are to apply the distribution of target molecular labels and the number of their occurrences to the two negative binomial distributions; the two negative binomial distributions are used to determine the number n of true molecular labels. Steps; include removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label occurs less than the nth abundant number of molecular labels. A molecular label having a number is included, and a true molecular label includes a molecular label having an occurrence number equal to or greater than the number of occurrences of the nth most abundant molecular label.

The present specification discloses a computer system for determining the number of targets. In some embodiments, the computer system comprises a computer-readable memory for storing executable instructions; one or more computer processors in contact with the computer-readable memory, wherein the one or more computer processors are (a). ) A step of attaching a probability bar code to a plurality of targets using a plurality of probability bar codes to generate a target with a plurality of probability bar codes, in which each of the plurality of probability bar codes includes a molecular label. (B) the step of acquiring sequencing data for a target with a probability bar code; (c) for one or more of the targets: (i) having an identifiable sequence associated with the target in the sequencing data. Using the steps of counting the number of molecular labels; (ii) identifying clusters of target molecular labels using directional proximity; and using the clusters of target molecular labels identified in (iii) (ii). Then, in the step of collapsing the sequencing data obtained in (b) and (iv) the step of estimating the number of targets, after the estimated number of targets is the step of folding the sequencing data in (ii). , A step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), is programmed by a viable instruction. Multiple targets include targets for the entire transcriptome of the cell.

In some embodiments, the target molecular labels within the cluster are within a predetermined directional proximity threshold of each other. The directional proximity threshold is a Hamming distance of 1. The target molecular label within the cluster comprises one or more parent molecule labels and a child molecule label of one or more parent molecule labels, wherein the number of occurrences of the parent molecule label is the number of occurrences of predetermined directional proximity. It is above the threshold. The predetermined number of directional proximity occurrence thresholds may be 2 × ( number of occurrences of child molecule labels) -1 .

In some embodiments, the step of collapsing the sequencing data obtained in (b) using the cluster of target molecular labels identified in (ii) converts the number of child molecule labels generated into the parent molecule label. Includes the process of attribution.

In some embodiments, the executable instruction can further program one or more computer processors to determine the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be 15-20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to carry out the statistical analysis are as follows: applying the distribution of target molecular labels and the number of their occurrences to two negative binomial distributions; using the two negative binomial distributions to determine the number n of true molecular labels. It comprises a step of determining; a step of removing the fake molecular label from the sequencing data obtained in (b), where the fake molecular label is more than the number of nth most abundant molecular labels generated . A molecular label having a low number of occurrences is included, and a true molecular label includes a molecular label having a number of occurrences equal to or higher than the number of occurrences of the nth most abundant molecular label.

The present specification discloses methods for correcting PCR or sequencing errors. In some embodiments, the method comprises the following: (a) the step of obtaining sequencing data for a probabilistic bar coded target; (b) for one or more of a plurality of targets: (i) in the sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target; (ii) a step of determining the number of noise molecular labels having an identifiable sequence associated with a target in sequencing data. And; (iii) including the step of estimating the number of targets, wherein the estimated number of targets is adjusted according to the number of noise molecule labels determined in (ii), (i). Correlates with the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in. In some embodiments, the method further comprises determining the sequencing status of the target in the sequencing data. The target sequencing status in the sequencing data is saturated sequencing, undersequencing, or oversequencing. In some embodiments, the method can be used to determine the number of targets. The method further comprises (c) using multiple probability barcodes to attach probability barcodes to multiple targets to generate multiple probability barcoded targets; (d) probability barcoded targets. Can include a step of sequencing to generate sequencing data for a target with a received probability barcode.

In some embodiments, the saturation sequencing status is determined by a target having a number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. The predetermined saturation threshold is about 6557 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The predetermined saturation threshold is about 65532 if the probability barcode contains about 65536 molecular labels with identifiable sequences. When the sequence status of the target in the sequencing data is the saturated sequencing status, the number of noise molecule labels determined in (ii) is zero.

In some embodiments, the undersequencing status can be determined by a target having a depth less than a predetermined undersequencing threshold (eg, mean, minimum, or maximum depth). The undersequencing threshold is about 4. The undersequencing threshold can be independent of the number of molecular labels with identifiable sequences. If the target sequencing status in the sequencing data is under-sequencing status, the number of noise molecule labels determined in (ii) is zero.

In some embodiments, the excess sequencing status is determined by a target comprising a number of molecular labels having an identifiable sequence that is greater than a predetermined excess sequencing threshold. For example, the excess sequencing threshold can be about 250 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The method determines the number of molecular labels having an identifiable sequence associated with a target in the sequencing data, given the predetermined excess sequence, when the sequencing data of the target in the sequencing data is in excess sequencing status. A step of subsampling to the single threshold is included.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in sequencing data is (iv) signal negative if the negative binomial distribution fitting condition is met. The step of fitting the binomial distribution to the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i), wherein the signal negative binomial distribution is a signal molecular label. There is a step corresponding to the number of molecular labels having an identifiable sequence associated with a target in the sequencing data counted in (i); (v) the noise-negative binomial distribution is counted in (i). A step of fitting to the number of molecular labels with identifiable sequences associated with a target in the sequenced data, where the noise-negative binomial distribution is the noise molecular label, the sequence counted in (i). The steps corresponding to the number of molecular labels with identifiable sequences associated with the target in the single data; the signal negative binomial distribution fitted in (vi) (v) and the noise negative binomial fitted in (vi). It comprises the steps of determining the number of noise molecule labels using the distribution.

In some embodiments, the negative binomial distribution fitting condition comprises that the target sequencing status in the sequencing data is not an under-sequencing status or an over-sequencing status. Using the signal-negative binomial distribution fitted in (v) and the noise-negative binomial distribution fitted in (vi), the step of determining the number of noise molecule labels is identifiable associated with the target in the sequencing data. For each of the molecular labels having an identifiable sequence, the step of determining that the signal probability of the identifiable sequence is a signal negative binomial distribution; and the noise probability of the identifiable sequence is the noise negative binomial distribution. A step of determining that the identifiable sequence is a noise molecule label if the signal probability is less than the noise probability.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in the sequencing data is that the sequencing status of the target in the sequencing data is under-sequencing status. Or if the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is less than the pseudo-point threshold, and not in excessive sequencing status, in (ii). Pseudo points to the number of molecular labels with identifiable sequences associated with the target in the sequencing data before determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. Includes the step of adding. The pseudo-point threshold is 10.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in the sequencing data is as follows: The sequencing status of the target in the sequencing data is undersequencing. If the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is greater than or equal to the pseudo-point threshold, not in single or excessive sequencing status. In ii), the step of removing the non-unique molecular label is included in determining the number of noise molecular labels having an identifiable sequence associated with the target in the sequencing data.

In some embodiments, the step of removing the non-unique molecular label is when the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than the predetermined reused molecular label threshold. In (ii), the step of removing the non-unique molecular label is included in determining the number of noise molecular labels having an identifiable sequence associated with the target in the sequencing data. For example, the reuse molecular label threshold can be about 650 if the probability barcode contains about 6651 molecular labels with identifiable sequences.

In some embodiments, the step of removing the non-unique molecular label determines the theoretical number of non-unique molecular labels for the number of molecular labels having an identifiable sequence associated with the target in the sequencing data. Steps; include removing the molecular label having a larger number of occurrences than the nth richest molecular label having an identifiable sequence associated with the target in the sequencing data, where n is: The theoretical number of non-unique molecular labels.

The present specification discloses a computer system for determining the number of targets. In some embodiments, the computer system comprises a computer-readable memory for storing executable instructions; one or more computer processors in contact with the computer-readable memory, wherein the one or more computer processors are (a). ) A step of attaching a probability bar code to a plurality of targets using a plurality of probability bar codes to generate a target with a plurality of probability bar codes, in which each of the plurality of probability bar codes includes a molecular label. (B) the step of acquiring sequencing data for a target with a probability bar code; (c) for one or more of the targets: (i) having an identifiable sequence associated with the target in the sequencing data. A step of counting the number of molecular labels; (ii) a step of determining the number of noise molecular labels having an identifiable sequence associated with a target in the sequencing data; and (ii) a step of estimating the number of targets. And, programmed by an executable instruction to perform, where the estimated number of targets was adjusted according to the number of noise molecule labels determined in (ii), counted in (i). Correlates with the number of molecular labels with identifiable sequences associated with the target in the sequenced data. In some embodiments, the method further comprises the step of determining the sequencing status of the target in the sequencing data. The target sequencing status in the sequencing data is saturated sequencing, undersequencing, or oversequencing.

In some embodiments, the saturation sequencing status is determined by a target having a number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. For example, a predetermined saturation threshold is about 6557 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The predetermined saturation threshold can be about 65532 if the probability barcode contains about 65536 molecular labels with identifiable sequences. When the sequence status of the target in the sequencing data is the saturated sequencing status, the number of noise molecule labels determined in (ii) is zero.

In some embodiments, the undersequencing status can be determined by a target having a depth less than a predetermined undersequencing threshold (eg, mean, minimum, or maximum depth). The undersequencing threshold is about 4. The undersequencing threshold can be independent of the number of molecular labels with identifiable sequences. If the target sequencing status in the sequencing data is under-sequencing status, the number of noise molecule labels determined in (ii) is zero.

In some embodiments, the excess sequencing status is determined by a target having a number of molecular labels with identifiable sequences that are greater than a predetermined excess sequencing threshold. For example, the excess sequencing threshold can be about 250 if the probability barcode contains about 6651 molecular labels with identifiable sequences. The method determines the number of molecular labels having an identifiable sequence associated with a target in the sequencing data, given that the sequence status of the target in the sequencing data is excessive sequencing. A step of subsampling to the single threshold is included.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in sequencing data is (iv) signal negative if the negative binomial distribution fitting condition is met. The step of fitting the binomial distribution to the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i), wherein the signal negative binomial distribution is a signal molecular label. There is a step corresponding to the number of molecular labels having an identifiable sequence associated with a target in the sequencing data counted in (i); (v) the noise-negative binomial distribution is counted in (i). A step of fitting to the number of molecular labels with identifiable sequences associated with a target in the sequenced data, where the noise-negative binomial distribution is the noise molecular label, the sequence counted in (i). The steps corresponding to the number of molecular labels with identifiable sequences associated with the target in the single data; the signal negative binomial distribution fitted in (vi) (v) and the noise negative binomial fitted in (vi). It comprises the steps of determining the number of noise molecule labels using the distribution.

In some embodiments, the negative binomial distribution fitting condition includes: the target sequencing status in the sequencing data is not an under-sequencing status or an over-sequencing status. Using the signal-negative binomial distribution fitted in (v) and the noise-negative binomial distribution fitted in (vi), the step of determining the number of noise molecule labels is identifiable associated with the target in the sequencing data. For each of the molecular labels having an identifiable sequence, the step of determining that the signal probability of the identifiable sequence is a signal negative binomial distribution; and the noise probability of the identifiable sequence is the noise negative binomial distribution. A step of determining that the identifiable sequence is a noise molecule label if the signal probability is less than the noise probability.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in the sequencing data is that the sequencing status of the target in the sequencing data is under-sequencing status. Or if the number of molecular labels with identifiable sequences associated with the target in the sequencing data counted in (i) is less than the pseudo-point threshold, and not in excessive sequencing status, in (ii). Pseudo points to the number of molecular labels with identifiable sequences associated with the target in the sequencing data before determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. Includes the step of adding. The pseudo-point threshold is 10.

In some embodiments, the step of determining the number of noise molecule labels with identifiable sequences associated with a target in the sequencing data is that the sequencing status of the target in the sequencing data is under-sequencing status. Or if the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i) is greater than or equal to the pseudo-point threshold, if not in excess sequencing status (ii). Includes the step of removing non-unique molecular labels in determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data.

In some embodiments, the step of removing the non-unique molecular label is when the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than the predetermined reused molecular label threshold. In (ii), the step of removing the non-unique molecular label is included in determining the number of noise molecular labels having an identifiable sequence associated with the target in the sequencing data. For example, the reuse molecular label threshold can be about 650 if the probability barcode contains about 6651 molecular labels with identifiable sequences.

The specification discloses one or more non-transient computer reading media, which, when performed, contain an executable code that implements any of the methods disclosed herein.

Shows non-limiting exemplary probability barcodes. Shows non-limiting exemplary probability barcoding and digital counting. FIG. 6 illustrates a non-limiting exemplary process for creating an indexed library of probabilistic barcode targets from multiple targets. It is a schematic diagram showing a non-limiting exemplary distribution of molecular labeling errors, sample labeling errors, and true molecular labeling signals. It is a flowchart which shows the non-limiting exemplary embodiment which corrects a PCR and a sequencing error using a molecular label. It is a schematic diagram which shows the sequencing data obtained by the complete sequencing and the incomplete sequencing. It is a flowchart showing a non-limiting exemplary embodiment of correcting PCR and sequencing errors using molecular labeling based on directional proximity. It is a flowchart showing a non-limiting exemplary embodiment which corrects PCR and sequencing error based on the quadratic derivative of recursive replacement error correction and molecular labeling depth change. FIG. 6 is a flow chart illustrating a non-limiting exemplary embodiment of correcting PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction. It is a flowchart which shows the non-limiting exemplary embodiment of error correction using two negative binomial distributions. It is a flowchart showing a non-limiting exemplary embodiment of correcting PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction by subsampling of microwell plates and mapping of molecular labels. It is a flowchart showing a non-limiting exemplary embodiment of correcting PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction by gene subsampling and molecular labeling mapping. It is a flowchart showing a non-limiting exemplary embodiment that corrects PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction by recursion. A flow chart illustrating a non-limiting exemplary embodiment that corrects PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction by using the second highest molecular label for initial parameter estimates. be. Shown are non-limiting exemplary devices suitable for use in the methods of the present disclosure. Shown are non-limiting exemplary structures of computer systems that can be used in connection with embodiments of the present disclosure. Illustrated is a non-limiting exemplary structure showing a network containing a plurality of computer systems suitable for use in the methods of the present disclosure. A non-limiting exemplary structure of a multiprocessor computer system using a shared virtual address memory space according to the methods of the present disclosure is shown. Non-limiting examples of complete and incomplete sequencing genes are shown. A non-limiting exemplary plot of corrected sequencing reads and their grades for single nucleotide sequencing errors as well as thresholds for separating true and error barcodes. It is a non-limiting exemplary figure of a zero-cut Poisson model. A bar graph of total sequencing leads per well is shown. Shown is a bar graph of complete sequencing genes (%), molecular labels (ML) (%) retained as true barcodes and retained leads (%) mapped to those MLs retained for each well. A boxplot of retained leads (%) that varies depending on the gene of each well is shown. Principal component analysis (PCA) using uncorrected ML vs. corrected MI after applying the algorithm from two plates is shown. It is an exemplary plot of the theoretical calculation of the unique molecular label used with the increase of input molecules. It is an exemplary plot showing the molecular label coverage of each molecular label across the microwell plate for the highly expressed gene-ATCB, where a clear distribution is observed between the error molecular label and the real molecular label. It is an exemplary plot showing the process of fitting the two negative binomial distributions to the molecular label coverage of each molecular label across the microwell plate for the highly expressed gene-ATCB. The fit of the two negative binomial distributions demonstrates that it is possible to statistically distinguish between a molecular labeling error with a lower molecular labeling depth and a true molecular labeling with a higher molecular labeling depth. The x-axis is the molecular depth. A molecular label correction was shown, where the pairwise Hamming distance of 1 accounted for a large proportion. After the molecular label was corrected, molecular labels with different Hamming distances of 1 were clustered and folded into the same parent molecule label. The curve of the number of corrected molecular labels to the corrected number of reads coverage is shown. A schematic diagram of an example of recursive replacement error correction is shown. Panels (a)-(e) show exemplary results with corrected PCR and sequencing errors based on the quadratic derivative of the molecular labeling depth change. Panels (a)-(c) show exemplary results with corrected PCR and sequencing errors based on two negative binomial distributions for CD69. Same as above. Panels (a)-(c) show exemplary results with corrected PCR and sequencing errors based on two negative binomial distributions for CD3E. Same as above. Panels (a)-(c) show exemplary results with corrected PCR and sequencing errors based on two negative binomial distributions for highly expressed genes. Same as above. Shown are exemplary results of the reuse of G-rich molecular labels for highly expressed genes. Panels (a)-(b) show exemplary results of adjusting input data for highly expressed genes prior to fitting the two negative binomial distributions. Panels (a)-(j) show a non-limiting exemplary validation of the dataset corrected using two negative binomial distributions. Same as above. Same as above. Same as above. Same as above. Panels (a)-(d) are exemplary t-probabilistic near-implantation (t) of Precise ™ targeting assays from 96 wells of mixed Jurkat and breast cancer (BrCa) single cells (86 test genes). -SNE) Shows visualization. Same as above. Panels (a)-(b) show unrestricted expression analysis of differences between cell clusters for genes> 0 ML in both selected clusters, calculated by DBScan and determined by the level of the genetic marker in each cluster. It is an exemplary plot. Same as above. Panels (a)-(d) are t-probabilistic neighborhood implants (t-probabilistic neighborhood implantation) of BD Precise ™ target assays from 96-well plates of mixed Jurkat and breast cancer (T47D) single cells containing 86 test genes. SNE) Non-limiting exemplary plot showing visualization. Same as above. Various identified in FIG. 41 before any error correction step (uncorrected ML shown in FIG. 42, panel (a)) and after RSEC and DBEC correction (adjusted ML shown in FIG. 42, panel (b)). It is a non-limiting exemplary heat map showing the difference gene expression by molecular labeling count between different cell clusters. Same as above.

In the following detailed description, the attached drawings that form a part thereof will be referred to. In these drawings, similar symbols generally consider similar components to be the same, unless context requires other interpretations. The detailed description, drawings, and exemplary embodiments described in the claims are not meant to be limiting. Other embodiments may be used or other modifications may be made without departing from the spirit or scope of the subject matter presented herein. As schematically described herein and illustrated in the drawings, aspects of the present disclosure can be arranged, substituted, combined, separated, and designed in a wide variety of different configurations, all of which can be arranged and designed. , It is to be understood that it is expressly considered herein and constitutes part of this disclosure.

All patents, published patent applications, other publications, and sequences from GenBank and other databases referenced herein are incorporated by reference in their entirety with respect to the relevant technology.

Quantification of a small number of nucleic acids, such as messenger ribonucleic acid (mRNA) molecules, is clinically important, for example, to determine genes that are expressed at different developmental stages or under different environmental conditions. However, it can be extremely difficult to determine the absolute number of nucleic acid molecules (eg, mRNA molecules), especially if the number of molecules is very small. One method of determining the absolute number of molecules in a sample is the digital polymerase chain reaction (PCR). Ideally, PCR produces the same copy of the molecule at each cycle. However, PCR can have problems such as amplification bias and inaccurate gene expression measurements because each molecule replicates with an putative probability, which varies with the PCR cycle and gene sequence. Probability barcodes with unique molecular labels (also called molecular indicators (MI)) can be used to count the number of molecules and correct for amplification bias. Probability bar coding, such as the Precise ™ assay (Cellular Research, Inc. (Palo Alto, CA)), PCR and PCR and by labeling mRNA during reverse transcription (RT) using molecular labeling (ML). The bias induced by the library fabrication process can be corrected.

The Precise ™ assay is labeled with a large number (eg, 6561-65536) on poly (T) oligonucleotides to hybridize to all poly (A) -mRNAs in the sample during the RT step. You can use a non-depleted pool of probability barcodes with. In addition to the molecular label, a sample label with a probability barcode (also referred to as sample index (SI)) can be used to identify each well of the Precise ™ plate. Probability barcodes may include universal PCR priming sites. During RT, the target gene molecule reacts randomly with the probability barcode. Each target molecule can hybridize with the resulting probability barcode to produce a complementary ribonucleotide acid (DNA) molecule with a probability barcode). After labeling, cDNA molecules with probability barcodes from the microwells of the microwell plate can be pooled in a single tube for PCR amplification and sequencing. Uncorrected sequencing data can be analyzed to obtain the number of reads, the number of probability barcodes with unique molecular labels, and the number of mRNA molecules according to Poisson correction or a correction method based on two negative binomial distributions.

Besides bias correction, molecular labeling can provide a better understanding of the statistical quality of the results by revealing the number of starting cDNA molecules present in the observed sequencing reads. For example, a large number of reads can give a statistically accurate answer, but measurement accuracy can be compromised if the reads are obtained from only a small number of starting mRNA molecules.

Amplification bias induced by PCR and library fabrication steps can be corrected, for example, by molecular labeling, but quantification of the absolute number of molecules can still be difficult due to some other factors. First, the estimation of the number of mRNA molecules can be limited by the overall variety of molecular labels. During probability barcoding, the mRNA molecule can react randomly with the available probability barcodes. Thus, although each mRNA molecule can hybridize to a probability barcode; its molecular label may not necessarily be unique for any given gene. When the number of mRNA molecules is smaller than the number of probability barcodes, each mRNA molecule tends to hybridize with a probability barcode with a unique molecular label, and the count of the number of molecules is the count of the number of molecular labels. Can be equivalent to.

As the number of mRNA molecules increases, a large number of mRNA molecules are more likely to hybridize to probability barcodes with the same molecular label. Therefore, the use of unique molecular label counts may underestimate the number of molecules. In some cases, the number of mRNA molecules can be estimated according to a Poisson correction or a correction based on two negative binomial distributions of the total number of unique molecular labels observed. However, in the extreme case where a complete collection of 6651 probability barcodes is observed, Poisson correction or correction based on two negative binomial distributions may no longer be possible. For example, regardless of whether the starting mRNA molecule is 65,000 or 100,000, the maximum value of the 6651 saturation probability barcode is expected in either case.

Second, PCR errors (ie, errors that occur during PCR amplification) can induce artificial probability barcodes and optionally increase the molecular label count. Third, PCR amplification bias and inefficient PCR can produce a small copy of the bar coded molecule that is indistinguishable from error. Fourth, sequencing errors, inaccurate calling of stochastic barcode sequences can induce artificial probabilistic barcodes and increase the molecular label count. In addition, sequencing depth can be important, especially if the sequencing is too shallow to detect all of the probabilistic barcoded mRNAs present in the sample library.

Methods and systems for determining the number of targets with one or more PCRs, or corrected or regulated sequencing errors, are disclosed herein. In some embodiments, the method is (a) a step of using a plurality of probability barcodes to attach probability barcodes to a plurality of targets to generate a plurality of targets with probability barcodes. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic bar code; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; and a step of determining the quality status of the target in the sequencing data obtained in (ii) (b); (iii). ) The step of determining one or more sequencing data errors in the sequencing data obtained in (b), wherein the step of determining one or more sequencing data errors in the sequencing data is as follows: Of the number of molecular labels with identifiable sequences associated with the target in the sequencing data, the quality status of the target in the sequencing data, and the number of molecular labels with identifiable sequences in multiple probability barcodes. A step comprising determining one or more; (iv) a step of estimating the number of targets, wherein the estimated number of targets results in one or more sequencing data errors determined in (iii). Containing a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), adjusted accordingly.

Disclosed are methods of determining the number of targets with one or more PCR or sequencing errors corrected or adjusted based on directional proximity. In some embodiments, the method is (a) a step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a plurality of targets with the probability barcodes. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic barcode; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; (iii) a step of identifying clusters of target molecular labels using directional proximity; (iii) (ii). ), Using the cluster of molecular labels of the targets identified in), the step of collapsing the sequencing data obtained in (b); (iv) the step of estimating the number of targets, the estimated number of targets. Includes (ii) folding of the sequencing data followed by a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i).

A computer system for determining the number of targets with one or more corrected or regulated PCR or sequencing errors is disclosed. Disclosed is a non-transient computer reading medium containing an executable code that, when executed, causes one or more computer devices to determine the number of targets with one or more corrected or regulated PCR or sequencing errors. Will be done.

Definitions Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. For example, Singleton et al. , Dictionary of Microbiology and Molecular Biology 2nd ed. , J. Wiley & Sons (New York, NY 1994); Sambrook et al. , Molecular Cloning, A Laboratory Manual, Cold Spring Harbor Press (Cold Spring Harbor, NY 1989). For the purposes of this disclosure, the following terms are defined below.

As used herein, the term "adapter" can mean a sequence for facilitating amplification or sequencing of related nucleic acids. The relevant nucleic acid may include the target nucleic acid. The relevant nucleic acid may include one or more of a spatial label, a target label, a sample label, an indicator label, a bar code, a stochastic bar code, or a molecular label. The adapter may be linear. The adapter may be a pre-adenylated adapter. The adapter may be double-stranded or single-stranded. One or more adapters can be placed at the 5'or 3'end of the nucleic acid. If the adapter contains known sequences at the 5'and 3'ends, the known sequences may be the same sequence or different sequences. Adapters located at the 5'and / or 3'ends of the polynucleotide may have the ability to hybridize to one or more oligonucleotides immobilized on the surface. The adapter, in some embodiments, comprises a universal array. The universal sequence may be one region of a nucleotide sequence common to two or more nucleic acid molecules. Two or more nucleic acid molecules can have regions of different sequences. Thus, for example, a 5'adapter may contain the same sequence and / or a universal nucleic acid sequence and a 3'adapter may contain the same sequence and / or a universal sequence. Universal sequences that can be present in different members of multiple nucleic acid molecules can allow replication or amplification of multiple different sequences using a single universal primer complementary to the universal sequence. Similarly, at least one, two (eg, a pair) or more universal sequences that may be present in different members of a collection of nucleic acid molecules are at least one, two (eg, a pair) that are complementary to the universal sequence. ) Or more single universal primers may be used to allow replication or amplification of multiple different sequences. Thus, universal primers include sequences that can hybridize to such universal sequences. The target nucleic acid sequence-bearing molecule can be modified to attach a universal adapter (eg, a non-target nucleic acid sequence) to one or both ends of a different target nucleic acid sequence. One or more universal primers bound to the target nucleic acid can provide a site for hybridization of the universal primer. The one or more universal primers bound to the target nucleic acid may be the same or different from each other.

As used herein, the terms "associated" or "associated with" can mean that at any given time two or more species can be identified as co-located. The association can mean that more than one species are in similar containers. The association can be an informatic association. In this case, for example, digital information about two or more species is stored, and the information can be used to determine that one or more of these species are co-located. The association can also be a physical association. In some embodiments, the two or more associated species are "tethered", "bonded", or "fixed" to the surface of a solid or semi-solid that is mutual or common. The association can mean covalent or non-covalent means for binding the label to a solid or semi-solid support such as beads. The association can be a covalent bond between the target and the label.

As used herein, the term "complementary" can mean the ability of precise pairing between two nucleotides. For example, if a nucleotide at a given position in a nucleic acid is hydrogen-bondable to a nucleotide in another nucleic acid, the two nucleic acids are considered complementary to each other at that position. Complementarity between two single-stranded nucleic acid molecules can be "partial" if only a portion of the nucleotide is bound, or complete if all of the single-stranded nucleic acid molecules have complementarity. It is possible. If the first nucleotide sequence is complementary to the second nucleotide sequence, then the first nucleotide sequence is said to be a "complement" of the second sequence. If the first nucleotide sequence is complementary to the reverse sequence of the second sequence (ie, the order of the nucleotides is reversed), then the first nucleotide sequence is said to be the "reverse complement" of the second sequence. .. As used herein, the terms "complementary", "complementary", and "reverse complementary" can be used synonymously. It is understood from the present disclosure that if one molecule can hybridize to another molecule, it can be a complement to the hybridizing molecule.

As used herein, the term "digital counting" can mean a method of estimating the number of target molecules in a sample. Digital counting may include determining the number of unique labels associated with the target in the sample. This probabilistic method transforms the problem of counting molecules from the problem of positioning and identifying the same molecule into a series of digital problems with and without the detection of a given set of labels.

As used herein, the term "label" can mean the nucleic acid code associated with a target in a sample. The label can be, for example, a nucleic acid label. The label can be a label that can be amplified in whole or in part. The label can be a sequenceable label in whole or in part. The label can be part of an individually identifiable natural nucleic acid. The label can be a known sequence. Labeling can include conjugation of nucleic acid sequences (eg, conjugation of natural and non-natural sequences). As used herein, the term "marker" may be used synonymously with the term "index," "tag," or "marker tag." Signs can convey information. For example, in various embodiments, the label can be used to determine sample identity, sample source, cell identity, and / or target.

As used herein, the term "non-depleted reservoir" can mean a pool of stochastic barcodes composed of a wide variety of labels. A non-depleted reservoir may contain a number of different probability barcodes such that if a non-depleted reservoir is associated with a pool of targets, each target is more likely to be associated with a unique probability barcode. The uniqueness of each labeled target molecule can be determined by random selection statistics and depends on the number of copies of the same target molecule in the collection compared to the variety of labels. The size of the resulting set of labeled target molecules can be determined by the probabilistic nature of the bar coding process, and then analysis of the number of probabilistic barcodes detected is performed on the target molecules present in the original collection or sample. Allows calculation of numbers. A labeled target molecule is highly unique (ie, the probability that two or more target molecules will be labeled with one given label) if the ratio of the number of copies of the existing target molecule to the number of unique probability barcodes is low. Is very low).

As used herein, the term "nucleic acid" means a polynucleotide sequence or fragment thereof. Nucleic acid can include nucleotides. Nucleic acids can be exogenous or endogenous to cells. Nucleic acid can be present in a cell-free environment. Nucleic acid can be a gene or a fragment thereof. Nucleic acid can be DNA. Nucleic acid can be RNA. Nucleic acids can include one or more analogs (eg, modified scaffolds, sugars or nucleic acid bases). Some examples of analogs are, but are not limited to, 5-bromouracil, peptide nucleic acid, xenonucleic acid, morpholino form, locked nucleic acid, glycolnucleic acid, treose nucleic acid, dideoxynucleotide, cordisepine, 7-deaza-GTP. , Fluorophore (eg, sugar-bound rhodamine or fluorescein), thiol-containing nucleotides, biotin-binding nucleotides, fluorescent base analogs, CpG islands, methyl-7-guanosine, methylated nucleotides, inosine, thiouridine, pseudouridine, dihydrouridine, Nucleic acid and Wyosin can be mentioned. "Nucleic acid", "polynucleotide", "target polynucleotide", and "target nucleic acid" can be used interchangeably.

Nucleic acids may include one or more modifications (eg, base modifications, skeletal modifications) to provide nucleic acids with new or improved characteristics (eg, improved stability). The nucleic acid may include a nucleic acid affinity tag. The nucleoside can be a base-sugar combination. The base portion of the nucleoside can be a heterocyclic base. The two most common classes of such heterocyclic bases are purines and pyrimidines. Nucleotides can be nucleosides that further contain a phosphate group covalently attached to the sugar moiety of the nucleoside. In nucleosides containing pentoflanosyl sugars, the phosphate group is capable of binding to the 2', 3', or 5'hydroxyl moieties of the sugar. When forming a nucleic acid, the phosphate group can covalently bond adjacent nucleosides to each other to form a linear polymer compound. As a result, the cyclic compound can be formed by further connecting the respective ends of the linear polymer compound. However, linear compounds are generally preferred. In addition, linear compounds can have internal nucleotide base complementarity and can be folded to produce fully double-stranded or partially double-stranded compounds. Within the nucleic acid, the phosphate group is usually referred to as forming the nucleoside interskeletal structure of the nucleic acid. The bond or backbone can be a 3'→ 5'phosphodiester bond.

Nucleic acids may contain modified scaffolds and / or modifications between modified nucleosides. Modified skeletons may include those that retain a phosphorus atom in the skeleton and those that do not have a phosphorus atom in the skeleton. Suitable modified nucleic acid skeletons containing a phosphorus atom include, for example, methyl such as phosphorothioate, chiral phosphorothioate, phosphorodithioate, phosphotriester, aminoalkylphosphotriester, 3'-alkylene phosphonate and 5'-alkylene phosphonate. And other alkyl phosphonates, chiral phosphonates, phosphinates, phosphoramidates such as 3'-aminophosphoramidate and aminoalkylphosphoramidates, phosphorodiamidates, thionophosphoramidates, thionoalkylphosphonates. , Thionoalkyl phosphotriesters, selenophosphates, and borane phosphates usually having 3'-5'bonds, 2'-5'binding analogs, and one or more internucleotide bonds 3'→ 3', 5'. It may include those having a reverse polarity, which is a → 5 ′ or 2 ′ → 2 ′ bond.

Nucleic acids are polynucleotide skeletons formed by short-chain alkyl or cycloalkyl nucleoside bonds, mixed heteroatoms and alkyl or cycloalkyl nucleoside bonds, or one or more short chain heteroatom or heterocycle nucleoside bonds. Can include. These include morpholino bonds (partially formed from the sugar moiety of the nucleoside), siloxane skeletons, sulfides, sulfoxides, and sulfone skeletons, form acetyl and thioform acetyl skeletons, methyleneform acetyl and thioform acetyl skeletons, riboacetyl skeletons, It may include alkene-containing skeletons, sulfamate skeletons, methyleneimino and methylenehydrazino skeletons, sulfonate and sulfoneamide skeletons, those with amide skeletons, and others with mixed N, O, S, and CH ₂ constituents.

Nucleic acids can include nucleic acid mimetics. The term "mimetic" may be intended to include a polynucleotide in which only the furanose ring or both the furanose ring and the internucleotide bond has been replaced with a non-furanose group, assuming that the replacement of the furanose ring alone is a sugar surrogate. It can be referred to. The heterocyclic base moiety or modified heterocyclic base moiety can be retained for hybridization with a suitable target nucleic acid. One such nucleic acid can be a peptide nucleic acid (PNA). In PNA, the sugar skeleton of a polynucleotide can be replaced with an amide-containing skeleton, particularly an aminoethylglycine skeleton. Nucleotides are retentive and are directly or indirectly attached to the azanitrogen atom of the amide portion of the backbone. The skeleton in the PNA compound may contain two or more bound aminoethylglycine units that give the PNA an amide-containing skeleton. The heterocyclic base moiety can be directly or indirectly attached to the azanitrogen atom of the amide moiety of the skeleton.

Nucleic acid may include a morpholino skeletal structure. For example, the nucleic acid may contain a 6-membered morpholino ring instead of a ribose ring. In some of these embodiments, the phosphodiester bond can be replaced by a phosphodiester bond or an internucleoside bond of another non-phosphodiester.

The nucleic acid may include a bound morpholino unit (ie, a morpholinonucleic acid) having a heterocyclic base bound to the morpholino ring. The binding group is capable of binding morpholinomonomer units in the morpholino nucleic acid. Nonionic morpholino oligomeric compounds may have less desirable interactions with cellular proteins. Morpholine polynucleotides can be nonionic mimics of nucleic acids. Various compounds within the morpholino class can be linked using different linking groups. A further class of polynucleotide mimetics can be referred to as cyclohexenyl nucleic acid (CeNA). The furanose ring normally present in nucleic acid molecules can be replaced with a cyclohexenyl ring. CeNA DMT-protected phosphoramidite monomers can be prepared and used for the synthesis of oligomeric compounds using phosphoramidite chemistry. Incorporation of CeNA monomers into the nucleic acid strand can increase the stability of the DNA / RNA hybrid. CeNA oligoadenylates can form complexes with nucleic acid complements that have stability similar to that of natural complexes. A further modification is a locked nucleic acid that forms a bicyclic sugar moiety by attaching a 2'-hydroxyl group to the 4'carbon atom of the sugar ring to form a 2'-C, 4'-C-oxymethylene bond. LNA) can be included. The bond can be a methylene (-CH2), group (where n is 1 or 2 in the formula) that crosslinks the 2'oxygen atom and the 4'carbon atom. LNA and LNA analogs may exhibit very high double-stranded thermal stability (Tm = + 3 to + 10 ° C.) with complementary nucleic acids, stability to 3'-exonuclease degradation, and good solubility.

Nucleic acids can also include modifications or substitutions of nucleobases (often referred to simply as "bases"). As used herein, "unmodified" or "natural" nucleobases are purine bases (eg, adenine (A) and guanine (G)), as well as pyrimidine bases (eg, thymine (T), cytosine (eg, thymine (T)). C) and uracil (U)) may be included. Modified nucleobases include other synthetic and naturally occurring nucleobases such as 5-methylcytosine (5-me-C), 5-hydroxymethylcytosine, xanthin, hypoxanthin, 2-aminoadenine, adenine and guanine 6-. Methyl and other alkyl derivatives, 2-propyl and other alkyl derivatives of adenine and guanine, 2-thiouracil, 2-thiothymine and 2-thiocytosine, 5-halolasyl and cytosine, 5-propynyl (-C = C-CH3) uracil And cytosine, and other alkynyl derivatives of pyrimidine base, 6-azouracil, cytosine and timine, 5-uracil (psoiduracil), 4-thiouracil, 8-halo, 8-amino, 8-thiol, 8-thioalkyl, 8-hydroxyl. , And other 8-substituted adenine and guanine, 5-halo, especially 5-bromo, 5-trifluoromethyl and other 5-substituted uracil and cytosine, 7-methylguanine and 7-methyladenine, 2-F-adenine, It may include 2-aminoadenine, 8-azaguanine and 8-azaadenine, 7-deazaguanine and 7-deazaadenine, and 3-deazaguanine and 3-deadenine. Modified nucleic acid bases include tricyclic pyrimidines, such as phenoxadincytidine (1H-pyrimid (5,4-b) (1,4) benzoxazine-2 (3H) -on), phenothiazinecytidine (1H-pyrimid) (5). , 4-b) (1,4) Benzoxazine-2 (3H) -on), substituted phenoxazine cytidine (eg 9- (2-aminoethoxy) -H-pyrimid (5,4- (b) (1, 4) G-clamps such as benzoxazine-2 (3H) -on), phenothiazine cytidine (1H-pyrimid (5,4-b) (1,4) benzothiadin-2 (3H) -on), substituted phenoxadin cytidine G-clamps such as (eg, 9- (2-aminoethoxy) -H-pyrimid (5,4- (b) (1,4) benzoxazine-2 (3H) -on), carbazole cytidine (2H-pyrimid). (4,5-b) indol-2-one), pyridoindole cytidine (H-pyrid (3',': 4,5) pyrrolo [2,3-d] pyrimidin-2-one) may be included.

As used herein, the term "sample" can mean a composition comprising a target. Suitable samples for analysis by the methods, devices, and systems of the present disclosure include cells, tissues, organs, or organisms.

As used herein, the term "sampling device" or "device" can mean a device capable of collecting and / or arranging sections on a substrate. The sample device can mean, for example, a fluorescence activated cell sorting (FACS) machine, a cell sorter machine, a biopsy needle, a biopsy device, a tissue sectioning device, a microfluidic device, a blade grid, and / or a microtome.

As used herein, the term "solid carrier" can mean a discrete solid or semi-solid surface to which multiple probability barcodes can be attached. Solid carriers are solid, porous, or hollow of any type composed of plastic, ceramic, metal, or polymeric materials (eg, hydrogels) capable of immobilizing nucleic acids (eg, covalently or non-covalently). It may include spheres, balls, bearings, cylinders, or other similar constructs. The solid support can be spherical (eg, microspheres) or discrete particles that can have irregular or irregular shapes, such as a cube, a rectangular parallelepiped, a pyramidal, a cylinder, a cone, an oblate, a disc, and the like. Can include. The plurality of solid supports arranged apart in an array may not contain a substrate. The solid carrier can be used synonymously with the term "beads".

The solid carrier can mean "base material". The substrate can be one of the solid carriers. Substrate can mean a continuous solid or semi-solid surface that can perform the methods of the present disclosure. The substrate can mean, for example, an array, a cartridge, a chip, a device, and a slide.

As used herein, the term "spatial marker" can mean a marker that can be associated with a position within space.

As used herein, the term "probability barcode" can mean a polynucleotide sequence containing a label. The probability barcode can be a polynucleotide sequence that can be used for probability barcoding. Probability barcodes can quantify targets in a sample. Probability barcodes can be used to control possible errors after associating a marker with a target. For example, probability barcodes can evaluate amplification or sequencing errors. The probability barcode associated with the target can be referred to as a probability barcode target or a probability barcode tag target.

As used herein, the term "gene-specific probability barcode" can mean a polynucleotide sequence comprising a label and a gene-specific target binding region. The probability barcode can be a polynucleotide sequence that can be used for probability barcoding. Probability barcodes can be used to quantify targets in a sample. Probability barcodes can be used to control possible errors after associating a marker with a target. For example, probability barcodes can evaluate amplification or sequencing errors. The probability barcode associated with the target can be referred to as a probability barcode target or a probability barcode tag target.

As used herein, the term "probability barcoding" can mean random labeling of nucleic acids (eg, barcoding). Probability bar coding can utilize a recursive Poisson strategy to associate a label with a target and quantify the label associated with the label. As used herein, the term "probability barcoding" may be used interchangeably with "gene-specific probability barcoding".

As used herein, the term "target" can mean a composition that can be associated with a probability barcode. Exemplary targets suitable for analysis by the methods, devices, and systems of the present disclosure include oligonucleotides, DNA, RNA, mRNA, microRNA, tRNA, and the like. The target can be single-stranded or double-stranded. In some embodiments, the target can be a protein. In some embodiments, the target is a lipid.

As used herein, the term "reverse transcriptase" can mean a group of enzymes that have reverse transcriptase activity (ie, catalyze the synthesis of DNA from an RNA template). In general, such enzymes are derived from, but not limited to, retrovirus reverse transcriptase, retrotransposon reverse transcriptase, retroplasma reverse transcriptase, letron reverse transcriptase, bacterial reverse transcriptase, and Group II intron. Included are reverse transcriptases and their variants, variants, or derivatives. Non-retroviral reverse transcriptase includes non-LTR retrotransposon reverse transcriptase, retroplasma reverse transcriptase, retron reverse transcriptase, and Group II intron reverse transcriptase. Examples of Group II intron reverse transcriptase include Lactococcus lactis Ll. The LtrB intron reverse transcriptase, Thermocynechococcus, extends the TeI4c intron reverse transcriptase, or Geobacillus stearomophilus GsI-IIC intron reverse transcriptase. Other classes of reverse transcriptase can include many classes of non-retroviral reverse transcriptases (ie, retrons, group II introns, and especially diversity-producing retroelements).

The terms "universal primer primer", "universal primer adapter" or "universal adapter sequence" are used interchangeably and are used to hybridize probability barcodes to generate gene-specific probability barcodes. Refers to a nucleotide sequence that can be. The universal adapter sequence may be, for example, a known sequence that is universal for all probability barcodes used in the methods of the present disclosure. For example, if multiple targets are labeled using the methods disclosed herein, each of the target-specific sequences may be linked to the same universal adapter sequence. In some embodiments, two or more universal adapter sequences can be used in the methods disclosed herein. For example, if multiple targets are labeled using the methods disclosed herein, at least two of the target-specific sequences are ligated with different universal adapter sequences. The universal adapter primer and its complement may be contained in two oligonucleotides, one containing a target-specific sequence and the other containing a probability barcode. For example, the universal adapter sequence may be part of an oligonucleotide that contains a target-specific sequence to generate a nucleotide sequence that is complementary to the target nucleic acid. A second oligonucleotide containing the probability barcode and the complementary sequence of the universal adapter sequence can hybridize with the nucleotide sequence to generate a target-specific probability barcode. In some embodiments, the universal adapter primer has a different sequence than the universal PCR primer used in the methods of the present disclosure.

The present specification discloses methods and systems for detecting and / or correcting errors that occur during PCR and / or sequencing. The types of errors include, but are not limited to, substitution errors (one or more bases) and unsubstituted errors. Of the substitution errors, single-base substitution errors can occur much more often than errors that differ by more than one base. The method and system can be used, for example, to achieve accurate counting of molecular targets by stochastic barcoding.

Probability Barcoding Probability barcoding is described, for example, in US Patent Application Publication No. 20150299784, International Publication No. 2015301691, and Fu et al, Proc Natl Acad Sci U.S.A. S. A. 2011 May 31; 108 (22): 9026-31, the contents of these publications are incorporated herein by reference in their entirety. Briefly, the probability barcode may be a polynucleotide sequence that can be used to attach a probability marker (eg, barcode, tag) to the target. Probability barcodes can include one or more indicators. Exemplary labels include universal labels, cell labels, molecular labels, sample labels, plate labels, spatial labels, and / or pre-spatial labels. FIG. 1 shows an exemplary probability bar code 104 with a spatial indicator. The probability barcode 104 may contain a 5'amine capable of linking the probability barcode to the solid support 105. Probability barcodes can include universal labels, dimensional labels, spatial labels, cellular labels, and / or molecular labels. The order of the various labels in the probability bar code, such as, but not limited to, universal, dimensional, spatial, cellular, and molecular labels can vary. For example, as shown in FIG. 1, the universal label may be the label on the most 5'side, and the molecular label may be the label on the most 3'side. Spatial, dimensional, and cellular labels may be in any order. In some embodiments, the universal label, spatial label, dimensional label, cell label, and molecular label may be in any order.

Labels, eg, cell labels, may contain a unique set of nucleic acid partial sequences of defined length, eg, 7 nucleotides each (corresponding to the number of bits used for some humming error correction codes). , Can be designed to provide error correction capability. A set of error-correcting subsequences comprises seven nucleotide sequences, which are designed so that any pair combination of sequences in the set exhibits a defined "gene distance" (or number of mismatched bases). And, for example, a set of error-correcting subsequences can be designed to exhibit a gene distance of 3 nucleotides. In this case, by reviewing the error-correcting sequences in the set of sequencing data for the labeled target nucleic acid molecule, amplification or sequencing errors can be detected or corrected. In some embodiments, the length of the nucleic acid partial sequence used to generate the error correction code is, for example, about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, ... It may be 20, 30, 31, 40, 50 nucleotides in length, or a number or range between any two of these values. In some embodiments, it is also possible to use nucleic acid partial sequences of other lengths to generate error correction codes.

The probability barcode may include a target binding region. The target binding region can interact with the target in the sample. Targets are ribonucleic acid (RNA), messenger RNA (mRNA), microRNA, small interfering RNA (siRNA), RNA degradation products, RNA each containing a poly (A) tail, or any combination thereof. Alternatively, these may be included. In some embodiments, the plurality of targets may include deoxyribonucleic acid (DNA).

In some embodiments, the target binding region may comprise an oligo (dT) sequence capable of interacting with the poly (A) tail of the mRNA. One or more of the probabilistic barcode labels (eg, universal, dimensional, spatial, cellular, and molecular labels) should be separated by a spacer from another one or two of the remaining probabilistic barcode labels. Can be done. Spacers are, for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 nucleotides or more. It may be there. In some embodiments, none of the probabilistic barcode markings are separated by spacers.

Universal Marking Probability barcodes can contain one or more universal markings. In some embodiments, one or more universal labels can be identical for all probability barcodes in a set of probability barcodes bound to a given solid carrier. In some embodiments, one or more universal labels can be identical for all probability barcodes bound to multiple beads. In some embodiments, the universal label may contain a nucleic acid sequence that is hybridizable to the sequencing primer. Sequencing primers can be used to sequence probability barcodes containing universal labels. The sequencing primer (eg, universal sequencing primer) may include a sequencing primer associated with a high throughput sequencing platform. In some embodiments, the universal label may include a nucleic acid sequence that is hybridizable to the PCR primer. In some embodiments, the universal label may contain a sequencing primer and a nucleic acid sequence that is hybridizable to the PCR primer. A universally labeled nucleic acid sequence capable of hybridizing to a sequencing primer or a PCR primer can be referred to as a primer binding site. Universal labels may contain sequences that can be used to initiate transcription of stochastic barcodes. Universal markings may contain sequences that can be used for probabilistic barcodes or extension of regions within probabilistic barcodes. Universal labels are approximately 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50 nucleotides in length, or in numbers or ranges between any two of these values. It may be there. For example, a universal label may contain at least about 10 nucleotides. The universal label can be at least, or at most, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 200, or 300 nucleotides in length. In some embodiments, the cleavable linker or modified nucleotide may be part of a universally labeled sequence that allows the probabilistic barcode to be cleaved and removed from the carrier.

Dimensional Marking Probability barcodes can contain one or more dimensional markings. In some embodiments, the dimension label may include a nucleic acid sequence that provides information about the dimension in which the probability labeling was performed. For example, a dimensional indicator can provide information about when a target is tagged with a probability bar code. Dimensional markers can be associated with the time point of probability barcoding of the sample. Dimensional labeling can be activated at the time of stochastic labeling. Different dimensional labels can be activated at different time points. Dimensional markers provide information about the order of targets, groups of targets, and / or samples with probability barcodes. For example, a cell population can attach a probability barcode to the G0 phase of the cell cycle. The cell can be pulsed again with a probability barcode during the G1 phase of the cell cycle. The cell can be pulsed again with a probability barcode during the S phase of the cell cycle, as well as at other periods. Probability barcodes at each pulse (eg, each phase of the cell cycle) can contain different dimensional markers. Thus, the dimensional labeling provides information about at what stage of the cell cycle the target was labeled. Dimensional markers can scrutinize a wide variety of biological times. Exemplary biological times include, but are not limited to, the cell cycle, transcription (eg, transcription initiation), and transcript degradation. As another example, it is possible to probabilistically label samples (eg, cells, cell populations) before and / or after drug treatment and / or therapy. Changes in the copy number of the identifiable target can be an indicator of the sample's response to the drug and / or therapy.

The dimensional label may be activating. The activating dimensional label can be activating at a particular point in time. Activateable labels can, for example, be constitutively activated (eg, do not switch off). The activating dimensional label is, for example, reversibly activating (eg, the activating dimensional label is switchable on and off). For example, the dimensional label can be reversibly activated at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 times or more. The dimensional label can be reversibly activated, for example, at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 times or more. In some embodiments, the dimensional label is fluorescent, light, chemical event (eg, cleavage, ligation of other molecules, modification (eg, pegging, SUMOization, acetylation, methylation, deacetylation, deacetylation). It can be activated by the addition of methylation), photochemical events (eg, photocaging), and the introduction of unnatural nucleotides.

The dimensional label can, in some embodiments, be the same for all probability barcodes attached to a given solid carrier (eg, beads), but can be different for different solid carriers (eg, beads). In some embodiments, at least 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%, or 100% of the probability barcodes on the same solid support are of the same dimension. May include signs. In some embodiments, at least 60% of the probability barcodes on the same solid support may contain the same dimensional label. In some embodiments, at least 95% of the probability barcodes on the same solid support may contain the same dimensional label.

Multiple solid carriers (eg, beads) can have as many as ¹⁰⁶ or more unique dimensionally labeled sequences. A dimensional indicator is a number or range between 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, or any two of these values, or an approximation thereof. The value can be the nucleotide length. The dimensional label can be at least, or at most, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 200, or 300 nucleotides in length. The dimensional label can contain from about 5 to about 200 nucleotides. The dimensional label may contain from about 10 to about 150 nucleotides. The dimensional label may contain from about 20 to about 125 nucleotides.

Spatial Marking Probability barcodes can contain one or more spatial markings. In some embodiments, the spatial label may include a nucleic acid sequence that provides information about the spatial orientation of the target molecule associated with the probability barcode. Spatial markers can be associated with coordinates in the sample. The coordinates can be fixed coordinates. For example, the coordinates can be fixed relative to the substrate. Spatial markers can be based on a two-dimensional or three-dimensional grid. The coordinates can be fixed with respect to the landmark. Landmarks can be identified in space. The landmark can be an imageable structure. Landmarks can be biological structures such as anatomical landmarks. Landmarks can be cellular landmarks (eg organelles). A landmark can be a non-natural landmark, eg, a structure with an identifiable identifier such as a color code, barcode, magnetism, fluorescence, radioactivity, or a unique size or shape. Spatial markers can be associated with physical partitions (eg, wells, containers, or droplets). In some embodiments, multiple spatial markers are used together to code one or more positions in space.

The spatial label may be the same for all probability barcodes bound to a given solid carrier (eg beads), but may be different for different solid carriers (eg beads). In some embodiments, the percentages of probability barcodes on the same solid carrier, including the same spatial label, are 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%. , 100 percent, or a number or range between any two of these values, or an approximation thereof. In some embodiments, the percentage of probability barcodes on the same solid carrier, including the same spatial label, is at least or at most 60%, 70%, 80%, 85%, 90%, 95%. , 97%, 99%, or 100%. In some embodiments, at least 60% of the probability barcodes on the same solid support may contain the same spatial label. In some embodiments, at least 95% of the probability barcodes on the same solid support may contain the same spatial label.

Multiple solid carriers (eg, beads) can have as many as ¹⁰⁶ or more unique spatially labeled sequences. Spatial markers are numbers or ranges between 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, or any two of these values, or an approximation thereof. The value can be the nucleotide length. Spatial labels can be at least or at most 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 200, or 300 nucleotides in length. Spatial labels can contain from about 5 to about 200 nucleotides. Spatial labels can contain from about 10 to about 150 nucleotides. Spatial labels can contain from about 20 to about 125 nucleotides.

Cell Labeling Probability barcodes can include one or more cell labels. In some embodiments, the cell label may include a nucleic acid sequence that provides information for determining which target nucleic acid is derived from which cell. In some embodiments, the cell label is the same for all probability barcodes bound to a given solid carrier (eg beads), but is different for different solid carriers (eg beads). In some embodiments, the percentage of probability barcodes on the same solid carrier, including the same cell label, is 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%. , 100 percent, or a number or range between any two of these values, or an approximation thereof. In some embodiments, the percentage of probability barcodes on the same solid carrier, including the same cell label, is 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%. , Or 100%, or an approximation thereof. For example, at least 60% of the probability barcodes on the same solid support may contain the same cell label. As another example, at least 95% of the probability barcodes on the same solid support may contain the same cell label.

Multiple solid carriers (eg, beads) can have as many as ¹⁰⁶ or more unique cell-labeled sequences. Cell labels are 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, or a number or range between any two of these values, or an approximation thereof. The value can be the nucleotide length. Cell labels can be at least, or at most, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 200, or 300 nucleotides in length. For example, a cell label can contain from about 5 to about 200 nucleotides. As another example, the cell label may contain from about 10 to about 150 nucleotides. As yet another example, the cell label may contain from about 20 to about 125 nucleotides.

Molecular Labeling Probability barcodes can include one or more molecular labels. In some embodiments, the molecular label may include a nucleic acid sequence that provides information for identifying a particular type of target nucleic acid species hybridized to a probability barcode. The molecular label may include a nucleic acid sequence that provides a counter to the specific presence of the target nucleic acid species hybridized to a probability barcode (eg, the target binding region).

In some embodiments, various sets of molecular labels are attached to a given solid carrier (eg, beads). In some embodiments, 10 ² , 10 ³ , 10 ⁴ , 10 ⁵ , 10 ⁶ , 10 ⁷ , 10 ⁸ , 10 ⁹ , or a number or range between any two of these values, or an approximation thereof. Unique molecularly labeled sequences can exist. For example, multiple probability barcodes may contain about 6651 molecular labels with identifiable sequences. As another example, multiple probability barcodes can include about 65536 molecular labels with identifiable sequences. In some embodiments, there may be at least, or at most, 10 ² , 10 ³ , 10 ⁴ , 10 ⁵ , 10 ⁶ , 10 ⁷ , 10 ⁸ or 10 ⁹ unique molecularly labeled sequences. The unique molecularly labeled sequence is attached to a given solid carrier (eg, beads).

The molecular label is 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50 nucleotides in length, or a number or range between any two of these values, or It can be such an approximate nucleotide length. The molecular label can be at least, or at most, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 200, or 300 nucleotides in length.

Target Binding Regions Probability barcodes can include one or more target binding regions. In some embodiments, the target binding region is capable of hybridizing to the target of interest. In some embodiments, the target binding region may comprise a target (eg, a target nucleic acid, a target molecule, eg, a cellular nucleic acid being analyzed), eg, a nucleic acid sequence that specifically hybridizes to a particular gene sequence. In some embodiments, the target binding region may comprise a nucleic acid sequence capable of binding (eg, hybridizing) to a particular position of a particular target nucleic acid. In some embodiments, the target binding region may comprise a nucleic acid sequence capable of specific hybridization to a restriction enzyme site overhang (eg, EcoRI attachment terminal overhang). The probability barcode can then be ligated to any nucleic acid molecule containing a sequence complementary to the restriction site overhang.

In some embodiments, the target binding region may comprise a non-specific target nucleic acid sequence. A non-specific target nucleic acid sequence can mean a sequence that can bind to a plurality of target nucleic acids independently of a specific sequence of the target nucleic acid. For example, the target binding region can contain a random multimer sequence or an oligo (dT) sequence that hybridizes to the poly (A) tail of the mRNA molecule. Random multimer sequences can be, for example, random dimers, random trimmers, random quatramers, random pentamers, random hexamers, random septamers, random octamers, random nonamars, random decamers, or higher orders of any length. Can be a random multimer sequence of. In some embodiments, the target binding region is identical for all probability barcodes bound to a given bead. In some embodiments, the target binding region of the plurality of probability barcodes bound to a given bead comprises two or more different target binding sequences. The target binding region can be a number or range of 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, or any two of these values, or a nucleotide length of such approximation. .. Or more or roughly at least such nucleotide lengths. The target binding region can be at most about 5, 10, 15, 20, 25, 30, 35, 40, 45, 50 nucleotides in length or longer.

In some embodiments, the target binding region may comprise an oligo (dT) capable of hybridizing to an mRNA containing a polyadenylation end. The target binding region can be gene-specific. For example, the target binding region can be configured to hybridize to a specific region of the target. Target binding regions are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23. , 24, 25, 26, 27, 28, 29, 30, or a number or range between any two of these values, or a nucleotide length of such an approximation. Target binding regions are at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20. , 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 nucleotides in length. The target binding region may be about 5-30 nucleotides in length. If the probability barcode contains a gene-specific target binding region, this probability barcode can be referred to as a gene-specific probability barcode.

Orientation Probability barcodes can include one or more orientations that can be used for alignment (eg, alignment) of probability barcodes. The probability barcode may include a portion for isoelectric focusing. Different probability barcodes can contain different isoelectric focusing points. When such a probabilistic barcode is introduced into the sample, the sample can perform isoelectric focusing to orient the probabilistic barcode into a known form. Thus, orientality can be used to create a known map of probability barcodes in the sample. Exemplary orientality may include electrophoretic mobility (eg, based on the size of a stochastic barcode), isoelectric point, spin, conductivity, and / or self-assembly. For example, probability barcodes that include self-assembly orientation can be self-assembled to a particular orientation upon activation (eg, nucleic acid nanostructures).

Affinity probability barcodes can contain one or more affinities. For example, spatial labels may include affinity. Affinities can include chemical and / or biological moieties that can facilitate the binding of stochastic barcodes to other entities (eg, cellular receptors). For example, affinity can include an antibody, eg, an antibody specific for a particular moiety (eg, a receptor) on a sample. In some embodiments, the antibody is capable of inducing a probability barcode into a particular cell type or molecule. Targets located in and / or in the vicinity of a particular cell type or molecule can be probabilistically labeled. In some embodiments, the affinity can provide spatial information in addition to the nucleotide sequence of the spatial label, as the antibody can direct the probability barcode to a particular position. The antibody may be a therapeutic antibody, such as a monoclonal antibody or a polyclonal antibody. The antibody may be humanized or chimeric. The antibody may be a naked antibody or a fusion antibody.

An antibody is a full-length (ie, naturally occurring or formed by a normal immunoglobulin gene fragment recombination process) immunoglobulin molecule (eg, an IgG antibody) or an immunoactive (ie, specific binding) portion of the immunoglobulin molecule, eg. It can be an antibody fragment.

The antibody fragment can be part of an antibody such as, for example, F (ab') 2, Fab', Fab, Fv, sFv. In some embodiments, the antibody fragment is capable of binding to the same antigen recognized by the full-length antibody. The antibody fragment is an isolated fragment consisting of a variable region of the antibody, for example, an "Fv" fragment consisting of a heavy chain and a light chain variable region and a recombination in which the light chain and the heavy chain variable region are connected by a peptide linker. It may contain a single chain polypeptide molecule (“scFv protein”). Exemplary antibodies include, but are not limited to, antibodies against cancer cells, antibodies against viruses, antibodies that bind to cell surface receptors (CD8, CD34, CD45), and therapeutic antibodies.

Universal Adapter Primers Probability barcodes can include one or more universal adapter primers. For example, gene-specific probability barcodes may include universal adapter primers. A universal adapter primer can mean a nucleotide sequence that is universal for all probability barcodes. Universal adapter primers can be used to construct gene-specific probability barcodes. Universal adapter primers are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23. , 24, 25, 26, 27, 28, 29, 30, or a number or range between any two of these values, or a nucleotide length of such an approximation. Universal adapter primers are at least or at most 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20. , 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 nucleotides in length. The universal adapter primer may be about 5-30 nucleotides in length.

Solid Carriers The probability barcodes disclosed herein can be combined with solid carriers in some embodiments. The solid carrier may be, for example, synthetic particles. In some embodiments, some or all of the molecular labels (eg, first molecular labels) of the plurality of probability barcodes (eg, first plurality of probability barcodes) on the solid support are at least one nucleotide. different. The cell labels of the probability barcodes on the same solid support may be the same. Cell labels for probability barcodes on different solid carriers can differ by at least one nucleotide. For example, the first cell label of the first plurality of probability barcodes on the first solid carrier may have the same sequence and the second plurality of probability barcodes on the second solid carrier may have the same sequence. The cell labels of 2 may have the same sequence. The first cell label of the first plurality of probability barcodes on the first solid carrier and the second cell label of the second plurality of probability barcodes on the second solid carrier are at least one nucleotide. It can be different. The cell label can be, for example, about 5-20 nucleotides in length. The molecular label can be, for example, about 5-20 nucleotides in length. The synthetic particles may be beads, for example.

The beads may be, for example, silica gel beads, regulated porous glass beads, magnetic beads, dyna beads, sephadex / sephadex beads, cellulose beads, polystyrene beads, or any combination thereof. Beads include polydimethylsiloxane (PDMS), polystyrene, glass, polypropylene, agarose, gelatin, hydrogel, paramagnetic materials, ceramics, plastics, glass, methylstyrene, acrylic polymers, titanium, latex, sepharose, cellulose, nylon, silicon, Or may include materials such as any combination thereof.

In some embodiments, the beads may be polymer beads, eg, deformable beads or gel beads, which are functionalized with probability barcodes (eg, from 10X Genomics (San Francisco, CA)). Gel beads etc.). In some embodiments, the gel beads may comprise a polymer-based gel. Gel beads can be made, for example, by encapsulating one or more polymer precursors in a droplet. Exposure of the polymer precursor to an accelerator (eg, tetramethylethylenediamine (TEMED)) can produce gel beads.

In some embodiments, the polymer beads can be dissolved, melted, or decomposed, for example, under desired conditions. The desired conditions may include environmental conditions. The desired conditions can cause dissolution, melting, or decomposition of the polymer beads in a controlled manner. Gel beads can be dissolved, melted, or decomposed by chemical, physical, biological, thermal, magnetic, electrical, optical, or any combination thereof.

For example, test substances and / or reagents such as oligonucleotide barcodes can enter through the diffusion of the inner surface of the gel beads (eg, oligonucleotide barcodes and / or materials used to make oligonucleotide barcodes). (Inside) and / or may be coupled / fixed to the outer surface of the gel beads or any other microcapsule described herein. Coupling / fixation may be mediated by any form of chemical bond (eg, covalent bond, ionic bond) or physical phenomenon (eg, van der Waals force, dipole-dipole interaction, etc.). In some embodiments, coupling / immobilization of the reagent to gel beads or any other microcapsule described herein includes, for example, an unstable moiety (eg, a chemical cross-linking agent described herein). It may be reversible, such as via a chemical cross-linking agent. When the stimulus is applied, the unstable part can be cleaved and the immobilized reagent can be released. In some cases, the unstable moiety is a disulfide bond. For example, if the oligonucleotide bar code is immobilized on the gel beads via a disulfide bond, the disulfide bond can be cleaved and the oligonucleotide bar code released from the beads by exposing the disulfide bond to a reducing agent. can. Unstable moieties may be contained as part of the gel beads or microcapsules, as part of the chemical linker that links the reagent or test material to the gel beads or microcapsules, and / or as part of the reagent or test material. good.

In some embodiments, the gel beads can include a wide variety of polymers, including but not limited to: polymers, heat sensitive polymers, photosensitive polymers, magnetic polymers, pH sensitive polymers, Salt-sensitive polymers, chemically sensitive polymers, polyelectrolytes, polysaccharides, peptides, proteins, and / or plastics. Polymers include, but are not limited to, poly (N-isopropylacrylamide) (PNIPAAm), poly (styrene sulfonate) (PSS), poly (allylamine) (PAAm), poly (acrylic acid) (PAA), poly. (Ethethyleneimine) (PEI), Poly (diallyldimethyl-ammonium chloride) (PDADMAC), Poly (pyrol) (PPy), Poly (vinylpyrrolidone) (PVPON), Poly (vinylpyridine) (PVP), Poly (methacrylic acid). ) (PMAA), poly (methylmethacrylate) (PMMA), polystyrene (PS), poly (tetrahydrofuran) (PTTH), poly (phthalaldehyde) (PTTH), poly (hexylviologen) (PHV), poly (L-lysine). ) (PLL), poly (L-arginine) (PARG), lactic acid-glycolic acid copolymer (PLGA) and the like.

Numerous chemical stimuli can be used to trigger bead destruction or degradation. Examples of these chemical changes include, but are not limited to, pH-mediated changes to the bead wall, bead wall disintegration through chemical cleavage of crosslinks, bead wall depolymerization triggers, and bead wall switching reactions. Can be mentioned. Bulk changes may also be used to trigger bead destruction.

Bulk or physical changes to microcapsules via various stimuli also provide many advantages in designing capsules to release reagents. Bulk or physical changes occur on a macroscopic scale, where bead breakage is the result of stimulus-induced mechanical physical forces. Such processes include, but are not limited to, pressure-induced fractures, bead wall melting, or bead wall porosity changes.

Biological stimuli can also be used to trigger the destruction or degradation of beads. In general, biological triggers are similar to chemical triggers, but in many cases biomolecules or molecules commonly present in survival systems such as enzymes, peptides, sugars, nucleic acids are used. For example, the beads may contain polymers with peptide crosslinks that are sensitive to cleavage by certain proteases. More specifically, one example may include microcapsules containing a GFLGK peptide crosslink. When a biological trigger such as the protease cathepsin B is applied, the peptide crosslinks in the shellwell are cleaved and the contents of the beads are released. In other cases, the protease may be thermally activated. In another example, the beads include a shell wall containing cellulose. The addition of the hydrolytic enzyme chitosan serves as a biological trigger for the cleavage of cellulose bonds, the depolymerization of shell walls, and the release of their internal contents.

In addition, the beads can be induced to release their contents upon application of thermal stimuli. Changes in temperature can cause a variety of changes in the beads. Changes in heat can cause the beads to melt, much like the bead walls collapse. In another case, heat can increase the internal pressure of the internal components of the bead so that the bead breaks or bursts. In yet another case, heat can transform the beads into a deflated, dehydrated state. In addition, heat can act on the heat sensitive polymer in the walls of the beads, causing the beads to break.

The inclusion of magnetic nanoparticles in the bead wall of the microcapsules can allow the bead to break and induce a large number of beads. The devices of the present disclosure may include magnetic beads for any purpose. In one example, the incorporation of Fe ₃ O ₄ nanoparticles into a polyelectrolyte-containing bead triggers fracture in the presence of vibrating magnetic field stimuli.

The beads can also be destroyed or decomposed as a result of electrical stimulation. Like the magnetic particles described in the previous section, electrically sensitive beads also enable bead rupture triggers, as well as other functions such as alignment, conductivity or redox reactions under electric fields. In one example, beads containing an electrically sensitive material are aligned under an electric field so that the release of internal reagents can be controlled. In another example, the electric field can also induce a redox reaction inside the bead wall itself, which can increase porosity.

The beads can also be destroyed using light stimulation. Numerous optical triggers are possible, including systems using various molecules such as nanoparticles and chromophores capable of absorbing photons of a particular range of wavelengths. For example, a metal oxide coating can be used as a capsule trigger. UV irradiation of the polyelectrolyte capsule coated with SiO ₂ can cause the bead wall to collapse. In yet another example, a photoswitch material such as an azobenzene group may be incorporated into the bead wall. When UV or visible light is applied, these chemicals undergo reversible cis-trans isomerization upon absorption of photons. In this aspect, the incorporation of a photon switch provides a bead wall that can collapse or become more porous upon application of the optical trigger.

For example, in the non-limiting example of the probability barcode shown in FIG. 2, after introducing cells such as single cells into multiple microwells of the microwell array at block 208, the beads are placed in the microwell array of block 212. Can be introduced into multiple microwells. Each microwell may contain one bead. The beads may contain multiple probability barcodes. The probability barcode may include a 5'amine region bound to the bead. Probability barcodes may include universal labels, molecular labels, target binding regions, or any combination thereof.

The probability barcodes disclosed herein can be associated (eg, bound) with a solid carrier (eg, beads). The probability barcode associated with the solid support may include a molecular label selected from the group containing at least 100 or 1000 molecular labels each having a unique sequence. In some embodiments, the different probability barcodes bound to the solid support may include molecular labels of different sequences. In some embodiments, a particular percentage of probability barcodes associated with a solid carrier comprises the same cell label. For example, the percentage is 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%, 100%, or a number or range between any two of these values, or so. It can be an approximation. As another example, the percentage can be at least or at most 60%, 70%, 80%, 85%, 90%, 95%, 97%, 99%, or 100%. In some embodiments, the probability barcode associated with the solid carrier may include the same cell label. Probability barcodes bound to different solid carriers may include different cell labels selected from the group containing at least 100 or 1000 cell labels with unique sequences.

The probability barcodes disclosed herein can be associated (eg, bound) with a solid carrier (eg, beads). In some embodiments, the step of attaching probabilistic barcodes to a plurality of targets in a sample can be performed using a solid carrier containing a plurality of synthetic particles bound to the plurality of probability barcodes. In some embodiments, the solid support may include a plurality of synthetic particles associated with a plurality of probability barcodes. Spatial labeling of multiple probability barcodes on different solid carriers can differ by at least one nucleotide. The solid support may include, for example, a plurality of two-dimensional or three-dimensional probability barcodes. The synthetic particles may be beads. The beads may be silica gel beads, regulated porous glass beads, magnetic beads, dyna beads, sephadex / sephadex beads, cellulose beads, polystyrene beads, or any combination thereof. The solid support may include polymers, matrices, hydrogels, needle array devices, antibodies, or any combination thereof. In some embodiments, the solid support may be floating. In some embodiments, the solid carrier can be embedded in a semi-solid or solid array. The probability barcode does not have to be bound to the solid support. Probability barcodes may be individual nucleotides. The probability barcode may be attached to the substrate.

As used herein, the terms "tethered", "bound", and "fixed" are used synonymously to covalently or non-covalently bind a stochastic barcode to a solid carrier. Can mean the means of. Any of a wide variety of different solid supports can be used as solid carriers for binding pre-synthesized probabilistic barcodes or for solid phase synthesis of probabilistic barcodes in situ.

In some embodiments, the solid support is beads. Beads are one or more types of solid, porous, or hollow spheres, balls, bearings, cylinders, or other similar constructs capable of immobilizing nucleic acids (eg, covalently or non-covalently). Can be included. The beads can be composed of, for example, plastic, ceramic, metal, or polymeric materials, or any combination thereof. The beads are or may contain discrete particles, which are spherical (eg, microspheres) or non-spherical or irregularly shaped, such as cubic, rectangular parallelepiped, pyramidal, cylindrical. It has a shape, a cone shape, a rectangular parallelepiped shape, a disc shape, and the like. In some embodiments, the beads can have a non-spherical shape.

Beads are, but are not limited to, paramagnetic materials (eg, magnesium, molybdenum, lithium, and tantalum), _ultranormal magnetic materials (eg, ferrite (Fe 3O ₄ , magnetite) nanoparticles), ferromagnetic materials (eg,). For example, iron, nickel, cobalt, some alloys of them, and some rare earth metal compounds), ceramics, plastics, glass, polystyrene, silica, methylstyrene, acrylic polymers, titanium, latex, sepharose, agarose, hydrogels, It can include various materials such as polymers, cellulose, nylon, and any combination thereof.

In some embodiments, the beads (eg, beads to which a probability barcode is attached) are hydrogel beads. In some embodiments, the beads comprise hydrogel.

Some embodiments disclosed herein include one or more particles (eg, beads). Each particle can contain multiple oligonucleotides (eg, probability barcodes). Each of the plurality of oligonucleotides may contain a molecularly labeled sequence, a cell labeled sequence, and a target binding region (eg, an oligo dT sequence, a gene-specific sequence, a random multimer, or a combination thereof). The cell-labeled sequence of each of the plurality of oligonucleotides may be the same. The cell-labeled sequences of oligonucleotides on different particles may differ so that oligonucleotides on different particles can be identified. The number of different cell-labeled sequences may differ in different implementations. In some embodiments, the number of cell-labeled sequences is 10,100,200,300,400,500,600,700,800,900,1000,2000,3000,4000,5000,6000,7000,8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 70000, 80000, 90000, 100000, 10 ⁶ , 10 ⁷ , 10 ⁸ , 10 ⁹ , or any number or range between two of these values, or It can be higher or an approximation thereof. In some embodiments, the number of cell-labeled sequences is at least, or at most 10, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, 6000. , 7000, 8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 70000, 80000, 90000, 100000, 10 ⁶ , ^{10 7} , 10 ⁸ or 109. In some embodiments, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200 of a plurality of particles. , 300, 400, 500, 600, 700, 800, 900, 1000 or less, or more contain oligonucleotides of the same cell sequence. In some embodiments, the plurality of particles containing oligonucleotides of the same cell sequence are at most 0.1%, 0.2%, 0.3%, 0.4%, 0.5%, 0.7. %, 0.8%, 0.9%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10% or more. In some embodiments, none of the plurality of particles contains the same cell-labeled sequence.

Multiple oligonucleotides of each particle may contain different molecularly labeled sequences. In some embodiments, the number of molecularly labeled sequences is 10,100,200,300,400,500,600,700,800,900,1000,2000,3000,4000,5000,6000,7000,8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 70000, 80000, 90000, 100000, 10 ⁶ , 10 ⁷ , 10 ⁸ , 10 ⁹ , or any number or range between two of these values, or It can be such an approximation. The number of molecularly labeled sequences is at least or at most 10, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, It can be 10000, 20000, 30000, 40,000, 50000, 60000, 70000, 80000, 90000, 100000, 10 ⁶ , 10 ⁷ , 10 ⁸ or 10 ⁹ . For example, at least 100 of a plurality of oligonucleotides contain different molecularly labeled sequences. As another example, in a single particle, at least 100, 500, 1000, 5000, 10000, 15000, 20000, 50000 of multiple oligonucleotides, a number or range between any two of these values, or more. , Contains different molecularly labeled sequences. Some embodiments provide a plurality of particles that include a probability barcode. In some embodiments, the ratio of molecularly labeled sequences that differ from the number (or copy or number) of targets generated is at least 1: 1, 1: 2, 1: 3, 1: 4, 1: 5, 1: 1. 6, 1: 7, 1: 8, 1: 9, 1:10, 1:11, 1:12, 1:13, 1:14, 1:15, 1:16, 1:17, 1:18, It can be 1:19, 1:20, 1:30, 1:40, 1:50, 1:60, 1:70, 1:80, 1:90, or more. In some embodiments, each of the plurality of oligonucleotides further comprises a sample label, a universal label, or both. The particles may be, for example, nanoparticles or microparticles.

The size of the beads can vary. For example, the diameter of the beads may range from 0.1 micrometer to 50 micrometer. In some embodiments, the diameter of the beads is 0.1, 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50 micrometers, Or it can be a number or range between any two of these values, or an approximation thereof.

The diameter of the beads can be related to the diameter of the wells of the substrate. In some embodiments, the diameter of the beads is 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100%, or these, than the diameter of the wells. It may be a number or range between any two of the values of, or a length as long or short by such an approximation. The diameter of the beads can be related to the diameter of the cell (eg, a single cell confined in a well of the substrate). In some embodiments, the diameter of the beads is 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100%, 200% of the diameter of the cells. , 250%, 300%, or a number or range between any two of these values, or may be as long or short as such an approximation.

The beads can be embedded and / or bonded to the substrate. The beads can be embedded and / or bonded to gels, hydrogels, polymers, and / or matrices. The spatial location of the beads within the substrate (eg, gel, matrix, scaffold, or polymer) can be identified using the spatial markings present on the probability barcode on the beads that can serve as location addresses.

Examples of beads include, but are not limited to, streptavidin beads, agarose beads, magnetic beads, Dynabed®, MACS® microbeads, antibody conjugated beads (eg, anti-immunoglobulin microbeads). ), Protein A-conjugated beads, Protein G-conjugated beads, Protein A / G-conjugated beads, Protein L-conjugated beads, Oligo (dT) -conjugated beads, Silica beads, Silica-like beads, Anti-biotin microbeads, Anti-fluorescence Dye microbeads and BcMag ™ carboxyl-terminated magnetic beads can be mentioned.

The beads can be associated (eg, impregnated) with quantum dots or fluorochromes to fluoresce in one fluorescent optical channel or multiple optical channels. Beads can be associated with iron oxide or chromium oxide to make them paramagnetic or ferromagnetic. The beads can be identifiable. For example, beads can be imaged using a camera. The beads may have a detectable code associated with the beads. For example, beads may include probability barcodes. Beads can change in size, for example, due to swelling in organic or inorganic solutions. The beads can be hydrophobic. The beads can be hydrophilic. The beads can be biocompatible.

Solid carriers (eg beads) are visible. The solid support may include a visualization tag (eg, a fluorescent dye). Solid carriers (eg beads) can be etched by identifier (eg number). The identifier can be visualized by imaging the beads.

Substrate and Microwell Array As used herein, substrate can mean a type of solid carrier. The substrate can mean a solid carrier that may contain the probability barcodes of the present disclosure. The substrate may include, for example, a plurality of microwells. For example, the substrate may be a well array containing two or more microwells. In some embodiments, the microwell may include a reaction chamber with a specified volume. In some embodiments, the microwell is capable of trapping one or more cells. In some embodiments, the microwell is capable of confining only one cell. In some embodiments, the microwell is capable of enclosing one or more solid carriers. In some embodiments, the microwell is capable of confining only one solid carrier. In some embodiments, the microwell encloses a single cell and a single solid support (eg, beads).

Probability Barcoding Methods The present disclosure provides a method for estimating the number of identifiable targets at identifiable locations in body samples (eg, tissues, organs, tumors, cells). The method involves placing a probabilistic barcode in close proximity to the sample, melting the sample, associating an identifiable target with the probabilistic barcode, amplifying the target, and / or digitalizing the target. It may include a step of counting. The method may further include the steps of analyzing and / or visualizing the information obtained from the spatial markings on the probability barcode. In some embodiments, one method comprises the step of visualizing a plurality of labels in a sample. The process of mapping multiple targets to a map of a sample may include the creation of a 2D or 3D map of the sample. A 2D or 3D map can be created before or after probabilistic barcodes are attached to multiple targets in the sample. The step of visualizing a plurality of targets in a sample may include the step of mapping the plurality of targets to the map of the sample. The step of mapping multiple targets to a sample map may include creating a two-dimensional or three-dimensional map of the sample. Two-dimensional and three-dimensional maps can be created before or after probabilistic barcodes are attached to multiple targets in the sample. In some embodiments, the 2D and 3D maps can be made before or after the sample is melted. The steps of melting the sample before or after making the 2D or 3D map include heating the sample, contacting the sample with detergent, changing the pH of the sample, or any combination thereof. Can include.

In some embodiments, the step of attaching a probability barcode to a plurality of targets comprises hybridizing the plurality of probability barcodes with the plurality of targets to produce a target with a probability barcode. The step of attaching a probability barcode to a plurality of targets may include a step of creating an indexed library of targets with a probability barcode. The step of creating an indexed library of targets with probability barcodes can be performed using a solid support containing a plurality of probability barcodes.

Contact of Samples and Probability Barcodes The present disclosure provides a method of contacting a sample (eg, a cell) with a substrate of the present disclosure. For example, a sample containing cells, organs, or tissue flakes can be contacted with a probability barcode. For example, a gravitational flow can bring the cells into contact, in which case the cells can precipitate to form a monolayer. The sample may be tissue flakes. The flakes can be placed on the substrate. The sample may be one-dimensional (eg, forming a planar surface). The sample (eg, cells) can be spread over the entire substrate, for example, by growing / culturing the cells on the substrate.

If the probability barcode is located in close proximity to the target, the target can hybridize with the probability barcode. Probability barcodes can be contacted in non-depleting proportions so that each of the identifiable targets can be coupled with the identifiable probability barcodes of the present disclosure. Targets can be cross-linked with probability barcodes to ensure efficient binding between targets and probability barcodes.

Cytolysis After distribution of cells and probability barcodes, the cells are lysable to release the target molecule. Cytolysis can be achieved by any of a variety of means, eg, by chemical or biochemical means, by osmotic shock, or by thermal lysis, mechanical lysis, or optical lysis. The cells can be a detergent (eg, SDS, Lidodecylsulfate, Triton X-100, Tween-20, or NP-40), an organic solvent (eg, methanol or acetone), or a digestive enzyme (eg, Proteinase K,). It can be lysed by the addition of a cell lysis buffer containing pepsin or trypsin), or any combination thereof. In order to improve the association between the target and the probability barcode, it is possible to change the diffusion rate of the target molecule, for example by lowering the temperature and / or increasing the viscosity of the lysate.

In some embodiments, the sample can be dissolved using filter paper. The filter paper can be dipped in the dissolution buffer on the filter paper. The filter paper is applicable to the sample at a pressure that can facilitate dissolution of the sample and hybridization of the sample's target to the substrate.

In some embodiments, the dissolution can be done by mechanical dissolution, thermal dissolution, optical dissolution, and / or chemical dissolution. Chemical lysis may include the use of digestive enzymes such as proteinase K, pepsin, trypsin. Dissolution can be performed by adding a dissolution buffer to the substrate. The lysis buffer may contain Tris HCl. The lysis buffer may contain at least about 0.01, 0.05, 0.1, 0.5, or 1 M or more of Tris HCl. The lysis buffer may contain at most about 0.01, 0.05, 0.1, 0.5, or 1 M or more of Tris HCl. The lysis buffer may contain approximately 0.1 M Tris HCl. The pH of the lysis buffer can be at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or higher. The pH of the lysis buffer can be at most about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or higher. In some embodiments, the pH of the lysis buffer is about 7.5. The lysis buffer may contain salts (eg LiCl). The concentration of salt in the lysis buffer can be at least about 0.1, 0.5, or 1 M or higher. The concentration of salt in the lysis buffer can be at most about 0.1, 0.5, or 1 M or more. In some embodiments, the concentration of salt in the lysis buffer is about 0.5 M. The lysis buffer may contain a surfactant (eg, SDS, Li-dodecylsulfate, Triton X, Tween, NP-40). The concentration of surfactant in the lysis buffer is at least about 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%, 0.05%, 0.1%, 0. It can be 5.5%, 1%, 2%, 3%, 4%, 5%, 6%, or 7% or more. The concentration of the surfactant in the lysis buffer is at most about 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%, 0.05%, 0.1%, It can be 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, or 7% or more. In some embodiments, the concentration of detergent in the lysis buffer is about 1% Li dodecyl sulfate. The time used for dissolution in this method may depend on the amount of detergent used. In some embodiments, the more surfactant used, the shorter the time required for dissolution. The lysis buffer may contain a chelating agent (eg, EDTA, EGTA). The concentration of chelating agent in the lysis buffer can be at least about 1, 5, 10, 15, 20, 25, or 30 mM or higher. The concentration of chelating agent in the lysis buffer can be at most about 1, 5, 10, 15, 20, 25, or 30 mM or higher. In some embodiments, the concentration of chelating agent in the lysis buffer is about 10 mM. The lysis buffer may contain a reducing reagent (eg, β-mercaptoethanol, DTT). The concentration of reducing reagent in the lysis buffer can be at least about 1, 5, 10, 15, 20 mM or higher. The concentration of reducing reagent in the lysis buffer can be at most about 1, 5, 10, 15, 20 mM or higher. In some embodiments, the concentration of reducing reagent in the lysis buffer is about 5 mM. In some embodiments, the lysis buffer may contain about 0.1 M Tris HCl, about pH 7.5, about 0.5 M LiCl, about 1% lithium dodecyl sulphate, about 10 mM EDTA, and about 5 mM DTT. ..

Melting can be done at a temperature of about 4, 10, 15, 20, 25, or 30 ° C. Dissolution can be performed for about 1, 5, 10, 15, or 20 minutes or longer. Lysating cells may contain at least about 100,000, 200,000, 300,000, 400,000, 500,000, 600,000, or 700,000 target nucleic acid molecules or more. Lysating cells can contain at most about 100,000, 200,000, 300,000, 400,000, 500,000, 600,000, or 700,000 target nucleic acid molecules or more.

Binding of Probability Barcode to Target Nucleic Acid Molecule After lysis of the cell and release of the nucleic acid molecule from it, the nucleic acid molecule can be randomly associated with the probability bar code of the co-localized solid carrier. The association may include hybridization of the target recognition region of the probability barcode to the complementary portion of the target nucleic acid molecule (eg, the oligo (dT) of the probability barcode may interact with the poly (A) tail of the target. be). The assay conditions used for hybridization (eg, buffer pH, ionic strength, temperature, etc.) can be selected to facilitate the formation of a particular stable hybrid. In some embodiments, nucleic acid molecules released from lysed cells can be associated with multiple probes on the substrate (eg, hybridize with probes on the substrate). When the probe contains an oligo (dT), the mRNA molecule can hybridize to the probe and be reverse transcribed. The oligo (dT) portion of the oligonucleotide can act as a primer for the first strand synthesis of the cDNA molecule. For example, in a non-limiting example of the probability barcode shown in FIG. 2, block 216, the mRNA molecule can hybridize the probability barcode on the beads. For example, single-stranded nucleotide fragments can hybridize to the target binding region of a stochastic barcode.

Binding may further include ligation of the target recognition region of the probability barcode with a portion of the target nucleic acid molecule. For example, the target binding region may contain a nucleic acid sequence that may allow specific hybridization to a restriction site overhang (eg, EcoRI attachment terminal overhang). The assay procedure may further include treating the target nucleic acid with a restriction enzyme (eg EcoRI) to produce a restriction site overhang. The probability barcode can then be ligated to any nucleic acid molecule containing a sequence complementary to the restriction site overhang. A ligase (eg, T4DNA ligase) can be used to ligate the two fragments.

For example, in the non-limiting example of the probability barcode illustrated in FIG. 2, block 220, labeled targets (eg, target-barcode molecules) from multiple cells (or multiple samples) are subsequently, eg, eg. Can be pooled in tubes. Labeled targets can be pooled, for example, by recovering beads to which probability barcodes and / or target-barcode molecules are bound.

Recovery of a solid support-based collection of bound target-barcode molecules can be achieved by the use of magnetic beads and an externally applied magnetic field. After pooling the target-barcode molecules, all further processing can proceed in a single reaction vessel. Further processing may include, for example, a reverse transcription reaction, an amplification reaction, a cleavage reaction, a dissociation reaction, and / or a nucleic acid extension reaction. Further processing reactions can be performed within the microwells, i.e., without first pooling the labeled target nucleic acid molecules of multiple cells.

Reverse Transcription The present disclosure provides a method of generating a stochastic target-barcode conjugate using reverse transcription (eg, in block 224 of FIG. 2). A stochastic target-barcode conjugate may include a probabilistic bar code and a complementary sequence of all or part of the target nucleic acid (ie, a cDNA molecule with a probabilistic bar code). Reverse transcription of the associated RNA molecule can occur by adding reverse transcriptase along with reverse transcriptase. The reverse transcription primer can be an oligo (dT) primer, a random hexanucleotide primer, or a target-specific oligonucleotide primer. Oligo (dT) primers can be 12-18 nucleotides in length, or approximately such nucleotides in length, and can bind to the endogenous poly (A) tail at the 3'end of mammalian mRNA. Random hexanucleotide primers can bind mRNA at various complementary sites. Target-specific oligonucleotide primers typically selectively prime the mRNA of interest.

In some embodiments, reverse transcription of the labeled RNA molecule can occur by the addition of reverse transcription primers. In some embodiments, the reverse transcription primer is an oligo (dT) primer, a random hexanucleotide primer, or a target-specific oligonucleotide primer. Generally, oligo (dT) primers are 12-18 nucleotides in length and bind to the endogenous poly (A) + tail at the 3'end of mammalian mRNA. Random hexanucleotide primers can bind mRNA at various complementary sites. Target-specific oligonucleotide primers typically selectively prime the mRNA of interest.

Reverse transcription can be repeated to generate multiple labeled cDNA molecules. The methods disclosed herein are at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 19. Alternatively, it may include a step of performing 20 reverse transcription reactions. The method may include at least about 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100 reverse transcription reactions.

Amplification Nucleic acid amplification reaction (eg, in block 228 of FIG. 2) can be performed one or more times to generate multiple copies of the labeled target nucleic acid molecule. Amplification may be performed by a multiplex method in which a plurality of target nucleic acid sequences are simultaneously amplified. The amplification reaction can be used to add a sequencing adapter to the nucleic acid molecule. The amplification reaction, if present, may include the step of amplifying at least a portion of the sample label. The amplification reaction may include the step of amplifying at least a portion of the cell and / or molecular label. The amplification reaction may include amplifying at least a portion of a sample tag, cell label, spatial label, molecular label, target nucleic acid, or a combination thereof. The amplification reaction was 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 15%, 20%, 25 of multiple nucleic acids. %, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97%, 100%, Alternatively, it may include the step of amplifying the range or number between any two of these values. The method may further include performing one or more cDNA synthesis reactions to generate one or more cDNA copies of the target-barcode molecule, including sample labels, cell labels, spatial labels, and / or molecular labels. ..

In some embodiments, the polymerase chain reaction (PCR) can be used to perform amplification. As used herein, PCR can mean a reaction that performs in vitro amplification of a particular DNA sequence by simultaneous primer extension of the complementary strand of DNA. As used herein, PCR refers to variants of the reaction, such as, but not limited to, RT-PCR, real-time PCR, nested PCR, quantitative PCR, multiplex PCR, digital PCR, and assembly PCR. Can be included.

Amplification of the labeled nucleic acid can include non-PCR-based methods. Examples of non-PCR-based methods include, but are not limited to, multiplex substitution amplification (MDA), transcription-mediated amplification (TMA), nucleic acid sequence-based amplification (NASBA), strand substitution amplification (SDA), real-time SDA. , Rolling circle amplification, or circle-circle amplification. Other non-PCR-based amplification methods include DNA-dependent RNA polymerase-driven RNA transcription amplification or multiple cycles of RNA-oriented DNA synthesis and transcription to amplify DNA or RNA targets, ligase linkage reaction (LCR), and Qβ replicase. (Qβ) method, use of parindrome probe, strand substitution amplification, oligonucleotide-driven amplification with restriction endonuclease, primer hybridized to nucleic acid sequence and resulting double strand cleaved prior to extension reaction and amplification Amplification methods such as strand substitution amplification using nucleic acid polymerase lacking 5'exonuclease activity, rolling circle amplification, and branch extension amplification (RAM). In some embodiments, amplification can produce a cyclized transcript.

In some embodiments, the method disclosed herein further comprises performing a polymerase chain reaction on a labeled nucleic acid (eg, labeled RNA, labeled DNA, labeled cDNA) to generate a probabilistic labeled amplicon. include. The labeled amplicon may be a double-stranded molecule. The double-stranded molecule can include a double-stranded RNA molecule, a double-stranded DNA molecule, or an RNA molecule hybridized to the DNA molecule. One or both strands of a double-stranded molecule may include a sample label, a spatial label, a cellular label, and / or a molecular label. The probability-labeled amplicon can be a single-stranded molecule. Single-stranded molecules can include DNA, RNA, or combinations thereof. The nucleic acids of the present disclosure may include synthetic or modified nucleic acids.

Amplification may include the use of one or more unnatural nucleotides. Unnatural nucleotides can include photolabile or triggering nucleotides. Examples of unnatural nucleotides may include, but are not limited to, peptide nucleic acids (PNAs), morpholino nucleic acids, and locked nucleic acids (LNAs), as well as glycol nucleic acids (GNA) and threose nucleic acids (TNAs). Unnatural nucleotides can be added in one or more cycles of the amplification reaction. Addition of unnatural nucleotides can be used to identify the product at a particular cycle or time point of the amplification reaction.

The step of performing the amplification reaction more than once may include the use of one or more primers. The one or more primers may contain, for example, at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 nucleotides or more. The one or more primers may contain at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 nucleotides or more. One or more primers may contain less than 12-15 nucleotides. One or more primers can anneal to at least some of the probability labeled targets. One or more primers can anneal to the 3'end or 5'end of multiple probability-labeled targets. One or more primers can anneal to the internal regions of multiple probability-labeled targets. The internal region is at least about 50, 100, 150, 200, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, from the 3'end of multiple probability-labeled targets. 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530, 540, 550, 560, 570, 580, 590, It can be 600, 650, 700, 750, 800, 850, 900, or 1000 nucleotides. One or more primers may include a constant panel of primers. The one or more primers may include at least one custom primer. One or more primers may include at least one control primer. One or more primers may include at least one gene-specific primer.

One or more primers may include universal primers. The universal primer can anneal to the universal primer binding site. The one or more custom primers can be annealed to a first sample label, a second sample label, a spatial label, a cell label, a molecular label, a target, or any combination thereof. One or more primers may include universal and custom primers. Custom primers can be designed to amplify one or more targets. The target can include a subset of all nucleic acids in one or more samples. The target may include a subset of all probability labeled targets in one or more samples. The one or more primers may include at least 96 custom primers or more. The one or more primers may include at least 960 custom primers or more. One or more primers may include at least 9600 custom primers or more. One or more custom primers can anneal to two or more different labeled nucleic acids. Two or more different labeled nucleic acids can correspond to one or more genes.

Any amplification scheme can be used in the methods of the present disclosure. For example, in one scheme, first round PCR can amplify molecules bound to beads using gene-specific primers and primers for one universal Illumina sequencing primer sequence. The second round of PCR can amplify the first PCR product using nested gene-specific primers flanked by Illumina sequencing primer 2 sequences and primers to the universal Illumina sequencing primer 1 sequence. The third round of PCR adds P5 and P7 and a sample index to make the PCR product an Illumina sequencing library. Sequencing with 150 bp × 2 sequencing can reveal cell and molecular labels on read 1, genes on read 2, and sample indexes on index 1 read.

In some embodiments, the nucleic acid can be removed from the substrate using chemical cleavage. For example, a chemical group or modified base present in a nucleic acid can be used to facilitate its removal from a solid carrier. For example, enzymes can be used to remove nucleic acids from the substrate. For example, nucleic acids can be removed from the substrate by restriction endonuclease digestion. For example, uracil-d-glycosylase (UDG) treatment of nucleic acids containing dUTP or ddUTP can be used to remove nucleic acids from the substrate. For example, nucleic acids can be removed from the substrate using enzymes that perform nucleotide excision, such as base excision repair enzymes, such as depurine / depyrimidine (AP) endonucleases. In some embodiments, the nucleic acid can be removed from the substrate using a photocleavable group and light. In some embodiments, the cleaving linker can be used to remove nucleic acid from the substrate. For example, the cleaving linker may include at least one of biotin / avidin, biotin / streptavidin, biotin / neutralvidin, Ig-protein A, photolabile linker, acid or base unstable linker group, or aptamer.

If the probe is gene-specific, the molecule can hybridize to the probe and be reverse transcribed and / or amplified. In some embodiments, amplification is possible after the nucleic acid has been synthesized (eg, after reverse transcription). Amplification can be performed by a multiplex method under the condition that a plurality of target nucleic acid sequences are simultaneously amplified. Amplification can add a sequencing adapter to the nucleic acid.

In some embodiments, amplification can be performed on the substrate using, for example, bridge amplification. It is possible to add a homopolymer tail to the cDNA to produce a terminal that was suitable for bridge amplification with an oligo (dT) probe on the substrate. In bridge amplification, the primer complementary to the 3'end of the template nucleic acid can be the first primer of each pair covalently attached to the solid particle. When a sample containing the template nucleic acid is in contact with the particles for a single thermal cycle, the template molecule is annealed to the first primer and the first primer is forwardly extended by the addition of nucleotides. It forms a double-stranded molecule consisting of a template molecule and a newly formed DNA strand complementary to the template. In the heating step of the next cycle, the double-stranded molecule is denatured to release the template molecule from the particle, leaving a complementary DNA strand attached to the particle via the first primer. In the annealing step of the subsequent annealing step, the complementary strand is capable of hybridizing to a second primer that is complementary to the segment of the complementary strand at the position removed from the first primer. This hybridization allows the complementary strand to form a bridge between the first and second primers that is covalently immobilized to the first primer and also to the second primer by hybridization. In the extension step, the second primer is extended in opposite directions by adding nucleotides in the same reaction mixture, thereby converting the bridge into a double-stranded bridge. The next cycle is then initiated and the double-stranded bridge is denatured with one end bound to the particle surface via the first and second primers, respectively, and the other end in the unbound state, respectively. It is possible to provide two single-stranded nucleic acid molecules with. In this second cycle of annealing and extension steps, each strand can hybridize to a previously unused additional complementary primer on the same particle to form a new single-stranded bridge. The two previously unused primers hybridized at this point can be extended to convert the two new bridges into double-stranded bridges.

Amplification reactions are at least 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 15%, 20%, 25%, 30% of multiple nucleic acids. , 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97%, or 100%. Can include.

Amplification of the labeled nucleic acid can include PCR-based or non-PCR-based methods. Amplification of the labeled nucleic acid may include exponential amplification of the labeled nucleic acid. Amplification of the labeled nucleic acid may include linear amplification of the labeled nucleic acid. Amplification can be performed by the polymerase chain reaction (PCR). PCR can mean a reaction that in vitro amplifies a particular DNA sequence by co-primer extension of the complementary strand of DNA. PCR includes variants of the reaction, such as, but not limited to, RT-PCR, real-time PCR, nested PCR, quantitative PCR, multiplex PCR, digital PCR, suppression PCR, semi-suppressive PCR, and assembly PCR. sell.

In some embodiments, amplification of the labeled nucleic acid comprises a non-PCR-based method. Examples of non-PCR-based methods include, but are not limited to, multiplex substitution amplification (MDA), transcription-mediated amplification (TMA), nucleic acid sequence-based amplification (NASBA), strand substitution amplification (SDA), real-time SDA. , Rolling circle amplification, or circle-circle amplification. Other non-PCR-based amplification methods include DNA-dependent RNA polymerase-driven RNA transcription amplification or multiple cycles of RNA-oriented DNA synthesis and transcription to amplify DNA or RNA targets, ligase linkage reaction (LCR), Qβ replicase ( Qβ), use of parindrome probe, strand substitution amplification, oligonucleotide driven amplification with limiting endonuclease, primer hybridized to nucleic acid sequence and resulting double strand cleaved prior to extension reaction and amplification Amplification methods include chain substitution amplification, rolling circle amplification, and / or branch extension amplification (RAM) using nucleic acid polymerases lacking 5'exonuclease activity.

In some embodiments, the methods disclosed herein further comprise the step of performing a nested polymerase chain reaction on an amplified amplicon (eg, a target). Amplicons can be double-stranded molecules. The double-stranded molecule can include a double-stranded RNA molecule, a double-stranded DNA molecule, or an RNA molecule hybridized to the DNA molecule. One or both strands of a double-stranded molecule may include a sample tag or a molecular identifier label. Alternatively, the amplicon can be a single-stranded molecule. Single-stranded molecules can include DNA, RNA, or combinations thereof. The nucleic acids of the invention may include synthetic or modified nucleic acids.

In some embodiments, the method comprises the step of repeatedly amplifying the labeled nucleic acid to produce a large number of amplicon. The methods disclosed herein are at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 19. Alternatively, it may include a step of performing an amplification reaction 20 times. Alternatively, the method involves performing at least about 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100 amplification reactions. include.

The amplification step may further include adding one or more control nucleic acids to one or more samples containing the plurality of nucleic acids. The amplification step may further include adding one or more control nucleic acids to the plurality of nucleic acids. The control nucleic acid may include a control label.

Amplification may include the use of one or more unnatural nucleotides. Unnatural nucleotides can include photolabile and / or triggering nucleotides. Examples of unnatural nucleotides include, but are not limited to, peptide nucleic acids (PNA), morpholino nucleic acids and locked nucleic acids (LNA), as well as glycol nucleic acids (GNA) and threose nucleic acids (TNA). Unnatural nucleotides can be added for more than one cycle of the amplification reaction. Addition of unnatural nucleotides can be used to identify the product at a particular cycle or time point of the amplification reaction.

The step of performing the amplification reaction more than once may include the use of one or more primers. One or more primers may contain one or more oligonucleotides. One or more oligonucleotides may contain at least about 7-9 nucleotides. One or more oligonucleotides may contain less than 12-15 nucleotides. One or more primers can anneal to at least a portion of the labeled nucleic acid. One or more primers can anneal to the 3'end and / or 5'end of multiple labeled nucleic acids. One or more primers can anneal to the internal regions of multiple labeled nucleic acids. The internal region is at least about 50, 100, 150, 200, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350 from the 3'end of the plurality of labeled nucleic acids. 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530, 540, 550, 560, 570, 580, 590, 600 , 650, 700, 750, 800, 850, 900, or 1000 nucleotides. One or more primers may include a constant panel of primers. The one or more primers may include at least one custom primer. One or more primers may include at least one control primer. One or more primers may include at least one housekeeping gene primer. One or more primers may include universal primers. The universal primer can anneal to the universal primer binding site. One or more custom primers can be annealed to a first sample tag, a second sample tag, a molecular identifier label, a nucleic acid, or a product thereof. One or more primers may include universal and custom primers. Custom primers can be designed to amplify one or more target nucleic acids. The target nucleic acid may include a subset of all nucleic acids in one or more samples. In some embodiments, the primer is a probe attached to the array of the present disclosure.

In some embodiments, the step of attaching a probability barcode to a plurality of targets in a sample further comprises creating an index index library of probabilistic barcoded fragments. Molecular labels with different probability barcodes may be different from each other. The step of making an index index library of targets with probability barcodes includes the step of making a plurality of index index polynucleotides from a plurality of targets in a sample. For example, in the case of a probabilistic bar code target index index library that includes a first index index target and a second index index target, the labeled region of the first index index polynucleotide is that of the second index index polynucleotide. Different from the labeled region by 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50 nucleotides, generally at least or at most these values, or of these values. The number or range of nucleotides between any two may differ. In some embodiments, the step of creating an index index library for a probabilistic bar coded target involves contacting a plurality of oligonucleotides, such as poly (T) and labeled regions, with a plurality of labels, such as mRNA molecules. A step of performing first-strand synthesis using a reverse transcription enzyme to generate a single-stranded labeled cDNA molecule, each containing a cDNA region and a labeled region, comprising a plurality of targets. Contains at least two mRNA molecules of different sequences, and the plurality of oligonucleotides comprises at least two oligonucleotides of different sequences. The steps to create an index index library for probabilistic barcoded targets are further to amplify single-stranded labeled cDNA molecules in order to generate double-stranded labeled cDNA molecules; and to generate labeled amplicons. , A step of performing nested PCR on a double-stranded labeled cDNA molecule. In some embodiments, the method may include making an adapter-labeled amplicon.

Probability barcoding can use nucleic acid barcodes or tags to label individual nucleic acid (eg, DNA or RNA) molecules. In some embodiments, this comprises adding DNA barcodes or tags to the cDNA molecule as they are generated from the mRNA. Nested PCR can perform PCR amplification bias minimization. Adapters can be added, for example, for sequencing using next-generation sequencing (NGS). Sequencing results can be used, for example, to sequence cell labels, molecular labels, and nucleotide fragments of one or more copies of the target located in block 232 of FIG.

FIG. 3 is a schematic diagram showing a non-limiting exemplary process for creating an index index library of probabilistic barcoded targets, eg mRNA. As shown in step 1, the reverse transcription process can encode each mRNA molecule containing a unique molecular label, a cell label, and a universal PCR site. In particular, probabilistic hybridization of the set of molecular identifier labels 310 with the poly (A) tail region 308 of the RNA molecule 302 can reverse-transcribe the RNA molecule 302 to produce the labeled cDNA molecule 304 containing the cDNA region 306. can. Each of the molecular identifier labels 310 may include a target binding region, eg, a poly (dT) region 312, a labeled region 314, and a universal PCR region 316.

In some embodiments, the cell label may contain 3-20 nucleotides. In some embodiments, the molecular label may contain 3-20 nucleotides. In some embodiments, each of the plurality of probability barcodes further comprises one or more universal labels and cell labels, the universal label is the same for the plurality of probability barcodes on a solid carrier, and the cell labels are. The same is true for multiple probability barcodes on solid carriers. In some embodiments, the universal label may contain 3-20 nucleotides. In some embodiments, the cell label comprises 3-20 nucleotides.

In some embodiments, the labeled region 314 may include a molecular label 318 and a cell label 320. In some embodiments, the labeled region 314 may include one or more universal labels, dimensional labels, and cell labels. The molecular label 318 is generally 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, even if it is 100 nucleotides in length. , At least, or at most, such nucleotide lengths, or may be a number or range of nucleotide lengths between any of these values. The cell label 320 is generally 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, even if it is 100 nucleotides in length. , At least, or at most, such nucleotide lengths, or may be a number or range of nucleotide lengths between any of these values. Universal labels are generally 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, even if they are 100 nucleotides in length. It may be at least or at most such nucleotide length, or it may be a number or range of nucleotide lengths between any of these values. The universal label may be the same for multiple probability barcodes on a solid carrier and the cell label may be the same for multiple probability barcodes on a solid carrier. Dimensional markers are generally 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, even if they are 100 nucleotides in length. It may be at least or at most such a nucleotide length, or it may be a number or range of nucleotide lengths between any of these values.

In some embodiments, the labeled region 314 is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, Contains 200, 300, 400, 500, 600, 700, 800, 900, 1000 different labels, generally contains different labels with such values, or at least or at most different labels with such values, or of these values. It may contain different numbers or ranges of signs between them. Each label is generally 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, even if it is 100 nucleotides in length. It may be at least or at most such nucleotide length, or it may be a number or range of nucleotide lengths between any of these values. The set of molecular identifier labels 310 is 10, 20, 40, 50, 70, 80, 90, 10 ² , 10 ³ , 10 ⁴ , 10 ⁵ , 10 ⁶ , 10 ⁷ , 10 ⁸ , 10 ⁹ , 10 ¹⁰ , 10 Includes ¹¹ , 10 ¹² , 10 ¹³ , 10 ¹⁴ , 10 ¹⁵ , 10 ²⁰ molecular identifier labels 310, or generally contains identifier labels 310 of such values, or at least, or at most, molecular identifier labels 310 of such values, or A number or range of molecular identifier labels 310 between any of these values may be included. Also, each set of molecular identifier labels 310 may include, for example, a unique labeled region 314. The labeled cDNA molecule 304 can be purified to remove the excess molecular identifier label 310. Purification may include Angle bead purification.

As shown in step 2, the product from the reverse transcription process of step 1 can be pooled in one tube and PCR amplified using the first PCR primer pool and the first universal PCR primer. The pooling step is possible with the unique labeled area 314. In particular, the labeled cDNA molecule 304 can be amplified to produce the nested PCR-labeled amplicon 322. Amplification may include multiplex PCR amplification. Amplification may include multiplex PCR amplification using 96 multiplex primers with a single reaction volume. In some embodiments, multiplex PCR amplification is 10, 20, 40, 50, 70, 80, 90, 10 ² , 10 ³ , 10 ⁴ , 10 ⁵ , 10 ⁶ , 10 ⁷ , 10 ⁸ with a single reaction volume. Use multiplex primers of 10 ⁹ , 10 ¹⁰ 10 ¹¹ 10 ¹² 10 ¹³ 10 ¹⁴ 10 ¹⁵ 10 ²⁰ or generally use multiplex primers of such values, at least or at most multiplex primers of such values. Multiplex primers can be used, or a number or range between any of these values. Amplification may include a first PCR primer pool 324 of custom primers 326A-C targeting a particular gene and universal primers 328. The custom primer 326 can hybridize to one region within the cDNA portion 306'of the labeled cDNA molecule 304. The universal primer 328 can hybridize to the universal PCR region 316 of the labeled cDNA molecule 304.

As shown in step 3 of FIG. 3, the product from the PCR amplification of step 2 can be amplified using a nested PCR primer pool and a second universal PCR primer. Nested PCR can minimize PCR amplification bias. In particular, the nested PCR-labeled amplicon 322 can also be further amplified by nested PCR. Nested PCR may include multiplex PCR containing Nested PCR primer pools 330 of Nested PCR primers 332a-c and a second universal PCR primer 328'in a single reaction volume. The nested PCR primer pool 328 is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400. , 500, 600, 700, 800, 900, 1000 different nested PCR primers 330, generally different nested PCR primers 330 with such values, or at least, or at most different nested PCR primers 330 with such values, or Nested PCR primers 330 may contain different numbers or ranges between any of these values. The nested PCR primer 332 contains an adapter 334 and can hybridize to one region within the cDNA portion 306'of the labeled amplicon 322. The universal primer 328'contains the adapter 336 and can hybridize to the universal PCR region 316 of the labeled amplicon 322. In this way, step 3 produces an adapter-labeled amplicon 338. In some embodiments, the nested PCR primer 332 and the second universal PCR primer 328' may not contain adapters 334 and 336. Alternatively, the adapters 334 and 336 can be ligated to the product of nested PCR to produce the adapter-labeled amplicon 338.

As shown in step 4, the PCR product from step 3 can be PCR amplified for sequencing with library amplification primers. In particular, the adapters 334 and 336 can be used to perform one or more more assays against the adapter-labeled amplicon 338. Adapters 334 and 336 can hybridize with primers 340 and 342. One or more primers 340 and 342 may be PCR amplification primers. The one or more primers 340 and 342 may be sequencing primers. One or more adapters 334 and 336 can be used for further amplification of the adapter labeled amplicon 338. One or more adapters 334 and 336 can be used for sequencing the adapter labeled amplicon 338. Primer 342 can contain a plate index index 344, whereby the amplicon produced using the same set of molecular identifier labels 318 is sequenced once using next generation sequencing (NGS). It can be sequenced by reaction.

Correction of PCR and Sequencing Errors The present specification discloses methods for determining the number of targets. In some embodiments, the method is (a) a step of using a plurality of probability barcodes to attach probability barcodes to a plurality of targets to generate a plurality of targets with probability barcodes. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic bar code; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; and a step of determining the quality status of the target in the sequencing data obtained in (ii) (b); (iii). ) The step of determining one or more sequencing data errors in the sequencing data obtained in (b), wherein the step of determining one or more sequencing data errors in the sequencing data is as follows: Of the number of molecular labels with identifiable sequences associated with the target in the sequencing data, the quality status of the target in the sequencing data, and the number of molecular labels with identifiable sequences in multiple probability barcodes. A step comprising determining one or more; (iv) a step of estimating the number of targets, wherein the estimated number of targets results in one or more sequencing data errors determined in (iii). Containing a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), adjusted accordingly. Steps (i), (ii), (iii), and (iv) can be performed for each of the plurality of targets. The method can be multiplexed.

In some embodiments, the method further comprises collapsing the sequencing data before determining one or more sequencing data errors. The step of collapsing the sequencing data comprises the step of assigning a copy of a target having a similar molecular label and having an occurrence number less than a predetermined folding occurrence number threshold as having the same molecular label for multiple targets. Here, the two copies of the target have similar molecular labels if the molecular labels of the two copies of the target differ by at least one base in the sequence.

The percentage of molecular labels in the sequencing data retained after adjusting the sequencing data in response to one or more sequencing data errors can vary. In some embodiments, the percentage of molecular labels in the sequencing data retained after adjusting the sequencing data in response to one or more sequencing data errors is 50%, 60%, 70%, 80%. , 90%, 95%, 99%, or 99.9%, or roughly such a percentage, or a number or range between any two of these values. In some embodiments, the percentage of molecular labels in the sequencing data retained after adjusting the sequencing data in response to one or more sequencing data errors is at least, or at most 50%, 60%. , 70%, 80%, 90%, 95%, 99%, or 99.9%.

Determining the Molecular Labeling Count FIG. 5 is a flow chart illustrating a non-limiting exemplary embodiment 500 that uses molecular labeling to correct PCR and sequencing errors. In the 500 embodiment, a plurality of probability barcodes are used to attach a probability barcode (each of the plurality of probability barcodes includes a molecular label) to the plurality of targets to generate a plurality of targets with probability barcodes. After the step and after the step of acquiring the sequencing data of the target with the probability barcode, it starts from the start block 504.

For a target, eg, a gene derived from a cell in a microwell of a microwell array, the number of molecular labels containing the identifiable sequence associated with the target in the sequencing data can be counted in block 508. In the sequencing data, the two copies of the target may have similar molecular labels, for example, the molecular labels of the two copies of the target may differ by one base in the sequence. The two copies of the target may both be true, one copy of the target may be true and the other copy of the target may be the result of sequencing or PCR errors, or the target. Both copies may be the result of sequencing or PCR errors.

Folding Blocks for Sequencing Data Block 512 allows you to fold the sequencing data. The step of collapsing the sequencing data comprises the step of attributing a copy of a target having a similar molecular label and having an incidence less than a predetermined folding incidence threshold as having the same molecular label for multiple targets. sell. The predetermined number of folding occurrence thresholds can vary from 1 to 100. In some embodiments, the predetermined folding occurrence threshold is 1, 2, 3, 4, 5, 6, 7, 8 if the probability barcode contains about 6651 molecular labels with identifiable sequences. , 9, 10, 17, 20, 30, 40, 50, 60, 70, 80, 90, 100, or generally such values, or a number or range between any two of these values. .. In some embodiments, the predetermined folding occurrence threshold is at least, or at most 1, 2, 3, 4, 5 where the probability barcode contains about 6651 molecular labels with identifiable sequences. , 6, 7, 8, 9, 10, 17, 20, 30, 40, 50, 60, 70, 80, 90, or 100. For example, the molecular label may be 8 nucleotides in length and each nucleotide position may be adenine (A), cytosine (C), guanine (G); C, G, thymine (T); A, G, T; or Since it may have three possibilities such as A, C, T, etc., it is possible to generate a unique molecular label of 38 = ⁶⁵⁶¹ .

In some embodiments, the predetermined folding occurrence threshold is 1, 2, 3, 4, 5, 6, 7, 8 if the probability barcode contains about 65536 molecular labels with identifiable sequences. , 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or generally such values, or a number or range between any two of these values. In some embodiments, a given folding occurrence threshold is at least, or at most, 1, 2, 3, 4, 5 if the probability barcode contains about 65536 molecular labels with identifiable sequences. , 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100. For example, the molecular label may be 8 nucleotides in length, and each nucleotide position may have ^four possibilities: A, C, G, T, thus producing a unique molecular label of 34 = 65536.

の分子標識を有するものであってよく、分子標識当たりのリードの数は、それぞれ、２６１、２、２、１、および１である。分子標識

は、それらが、分子標識ＴＧＴＧＣＧＴＧと１ヌクレオチド（下線部）異なっているため、分子標識ＴＧＴＧＣＧＴＧと類似している。識別可能な配列を有する６５６１の分子標識があり、かつ所定の折りたたみ発生数閾値が７である場合、分子標識

の発生数は、分子標識ＴＧＴＧＣＧＴＧに帰属させることができる。 For example, there can be five copies of the target. Five copies of the target

The number of leads per molecular label may be 261, 2, 2, 1, and 1, respectively. Molecular labeling

Is similar to the molecularly labeled TGTGCGTG because they differ by one nucleotide (underlined) from the molecularly labeled TGTGCGTG. If there is a molecular label of 6561 with an identifiable sequence and the predetermined number of folds threshold is 7, then the molecular label

The number of occurrences of can be attributed to the molecular label TGTGCGTG.

の分子標識を有するものであってよく、分子標識当たりのリードの数は、それぞれ、１０、７、５、４、１、１、および１である。分子標識

は、分子標識ＣＧＣＧＴＴＣＡと、互いに１ヌクレオチド（下線部）異なっているため、類似している。識別可能な配列を有する６５６１の分子標識があり、かつ所定の折りたたみ発生数閾値が７である場合、分子標識

の発生数は、分子標識ＣＧＣＧＴＴＣＡに帰属させることができる。 As another example, there may be seven copies of the target. Seven copies of the target

The number of leads per molecular label may be 10, 7, 5, 4, 1, 1, and 1, respectively. Molecular labeling

Is similar to the molecularly labeled CGCGTTCA because they differ by 1 nucleotide (underlined) from each other. If there is a molecular label of 6561 with an identifiable sequence and the predetermined number of folds threshold is 7, then the molecular label

The number of occurrences of can be attributed to the molecular label CGCGTTCA.

Sequencing Data Errors The methods disclosed herein can be used to identify and / or correct sequencing data errors, eg, errors that occur in the method of counting one or more target nucleic acids. In some embodiments, the sequencing data error may include, or may be, a PCR introduction error, a sequencing introduction error, an error due to bar code contamination, a library fabrication error, or any combination thereof. PCR induction errors may include or may be the result of PCR amplification errors, PCR amplification bias, inadequate PCR amplification, or any combination thereof. Sequencing implementation errors may include or may result from inaccurate base calling, inadequate sequencing, or any combination thereof. The error may include, or may be, a deletion of one or more nucleotides, a substitution of one or more nucleotides, an addition of one or more nucleotides, or any combination thereof.

Determining Sequencing Status As mentioned above, by using multiple probability barcodes and attaching probability barcodes to multiple targets, it is possible to generate multiple targets with probability barcodes, and multiple probability barcodes can be generated. Each of these may include molecular labeling, as well as acquisition of sequencing data for targets with probability barcodes. For labels, eg, genes derived from one cell within a microwell of a microwell array, the number of molecular labels with identifiable sequences associated with the target in the sequencing data can be counted. The counted sequencing data, for example, assigns a copy of a target having a similar molecular label and having an incidence less than a predetermined folding incidence threshold as having the same molecular label for multiple targets. It can be folded depending on the process. After collapsing the sequencing data, the quality status of the target can be determined.

With reference to FIG. 5, in some embodiments, the quality status of the target in block 516, sequencing data can be determined to be complete sequencing, incomplete sequencing, or saturated sequencing. The quality status of the target may depend on whether all of the true or real molecular labels were observed at the depth of the sequencing run. A true or real molecular label can mean a molecular label that is not an error or false molecular label. An error or false molecular label can mean a molecular label having a sequence resulting from a PCR error, man-made object, or sequencing error. The quality status of a target in the sequencing data is the number of molecular labels with an identifiable sequence in multiple probability barcodes and the molecule with an identifiable sequence associated with the target in the counted sequencing data. It can be determined by the number of signs.

In some embodiments, the complete sequencing quality status can be determined by a dispersal index compared to a Poisson distribution above a predetermined complete sequencing dispersal threshold. The dispersal index can be defined as the variance / average of the targets. FIG. 6 is a schematic diagram showing sequencing data obtained by complete sequencing and incomplete sequencing. FIG. 6 shows three copies of gene A and six copies of gene B in the library (left circle). If three copies of gene A had six, five, and one sequencing reads in the sequencing data (upper right circle), the variance was 7, the mean was 4, and the dispersal index was 1.75. Is. If the six copies of gene B had nine, two, two, two, one, and one sequencing reads in the sequencing data (upper right circle), the variance was 9.36. The average is 2.83 and the dispersion index is 3.31. Using these sequencing data, Gene A and Gene B can be considered to have complete sequencing status if a given complete sequencing application threshold is, for example, 0.9 for complete sequencing.

If one copy of gene A is not observed and the other two copies of gene A have two and three sequencing reads in the sequencing data (bottom right circle), the variance is 0.5. The average is 2.5 and the dispersion index is 0.2. Two copies of gene B were not observed, and the other four copies of gene B had four, two, one, and one sequencing reads in the sequencing data (lower right circle). In the case, the variance is 2, the average is 2, and the dispersal index is 2. Using these sequencing data, Gene A and Gene B can be considered to have incomplete sequencing status if a given complete sequencing application threshold is, for example, 1.1 for complete sequencing. ..

A given complete sequencing application threshold can vary from 0.5 to 5. In some embodiments, a given complete sequencing application threshold is 0.5, 0.6, 0.7, 0.8, 0.9, 1, 2, 3, 4, 5, 6, or roughly. It can be such a value, or it can be a number or range between any two of these values. In some embodiments, a given complete sequencing application threshold is at least or at most 0.5, 0.6, 0.7, 0.8, 0.9, 1, 2, 3, 4, 5 , Or it can be 6.

In some embodiments, the complete sequencing quality status can also be further determined by a molecular label having an incidence greater than or equal to a predetermined complete sequencing occurrence threshold in the sequencing data. The predetermined complete sequencing occurrence threshold can vary from 8 to 20. In some embodiments, the complete sequencing occurrence threshold is 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or generally such a value, or It can be a number or range between any two of these values. In some embodiments, the complete sequencing occurrence threshold can be at least, or at most 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20. ..

In some embodiments, the saturation sequencing quality status can be determined by a target having a number of molecular labels containing an identifiable sequence that is greater than a predetermined saturation threshold. Saturation sequencing quality status can also be further determined by the other one of a plurality of targets having a number of molecular labels containing identifiable sequences greater than a predetermined saturation threshold.

A given saturation threshold can fluctuate. In some embodiments, the predetermined saturation threshold is 6000, 6100, 6200, 6300, 6400, 6500, 6557, 6558, 6559 if the probability barcode contains about 6651 molecular labels with identifiable sequences. , 6560, 6651, or generally such values, or can be a number or range between any two of these values. In some embodiments, the predetermined saturation threshold is at least, or at most, 6000, 6100, 6200, 6300, 6400, 6500 if the probability barcode contains about 6651 molecular labels with identifiable sequences. , 6557, 6558, 6559, 6560, or 6651. In some embodiments, a given saturation threshold is 64000, 64100, 64200, 64300, 64400, 64500, 64600, 64700, 64800 if the probability bar code contains about 65536 molecular labels with identifiable sequences. , 64900, 65000, 65100, 65200, 65300, 65400, 65500, 65510, 65520, 65530, 65532, 65533, 65534, 65535, or generally such values, or a number between any two of these values. It can be a range. In some embodiments, the predetermined saturation threshold is at least, or at most, 64000, 64100, 64200, 64300, 64400, 64500, where the probability bar code contains about 65536 molecular labels with identifiable sequences. , 64600, 64700, 64800, 64900, 65000, 65100, 65200, 65300, 65400, 65500, 65510, 65520, 65530, 65532, 65533, 65534, or 65535.

In some embodiments, the quality status of the target in the sequencing data is classified as incomplete sequencing if the quality status of the target in the sequencing data is not complete sequencing and not saturated sequencing. can do.

Full Sequencing Quality Status The methods disclosed herein can provide an estimate of the number of targets in a sequencing library if the target has a complete sequencing quality status. If the targets in the sequencing library have full sequencing quality status, thresholds can be established for sequencing reads of true and error probability barcodes via individual Poisson models. The quality status of the target can depend on whether all of the true or real molecular labels were observed at the depth of the sequencing run. A true or real molecular label can mean a molecular label that is not an error or false molecular label. An error or false molecular label can mean a molecular label having a sequence resulting from a PCR error, man-made object, or sequencing error.

With reference to FIG. 5, in the determined state 520, if the target molecule has a complete sequencing status, embodiment 500 proceeds to block 524. In block 524, the single base sequencing error can be removed by the following steps. Step (1) If the sequencing lead is greater than 25, the molecular label associated with the most abundant sequencing lead is selected as the first parent molecular label. For example, after counting the number of molecular labels with identifiable sequences associated with the target in the sequencing data, the molecular label associated with the target in the sequencing data with the highest sequencing read is selected.

Step (2), Child Molecular Label: Identify a molecular label that has a sequencing read ≤ 3 and is one base away from the first parent molecule label; if no child molecule label or one-base child molecule label is found. Proceed to step (5). Step (3), multiple binomial tests are performed on all child molecule labels and parent molecule labels, the child molecule labels for which the null hypothesis is accepted are removed, and then their sequencing leads are subjected to them. Attributable to the parent of. If none of the null hypotheses are tolerated, this means that not all child molecule labels are single-base sequencing errors of the parent molecule label, in which case no read correction needs to be performed. Step (4), molecularly labeled sequences and sequencing reads are updated. For example, if the null hypothesis of multiple binomial tests is accepted, the number of child molecule labels generated can be attributed to the parent molecule label. Step (5), the molecular label with the next largest sequencing lead is selected as the parent molecule label and the above steps are repeated until there are no eligible parent molecule labels or eligible child molecule labels.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data is one or more if the target has a complete sequencing quality status. The step of determining all child molecule labels for the parent molecule label of the cell; and the step of performing a statistical analysis such as multiple binomial tests on at least one child molecule label and the parent molecule label; If so, the number of child molecule labels generated can be adjusted by the step of assigning to the parent molecule label.

In some embodiments, the child molecule label may comprise a molecule label that is one base different from the parent molecule label and has a number of occurrences that is less than or equal to a given complete sequencing child threshold. A given complete sequencing child threshold can fluctuate. In some embodiments, the predetermined complete sequencing child threshold is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or roughly such, or any of these values. It can be a number or range between the two. In some embodiments, the predetermined complete sequencing child threshold can be at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10.

In some embodiments, one or more parent molecular labels include a molecular label having an incidence greater than or equal to a predetermined complete sequencing parent threshold, wherein the predetermined complete sequencing parent threshold is a predetermined complete sequence. Equal to a single occurrence threshold, for example 8. The null hypothesis of the first statistical analysis is acceptable if the probability that the null hypothesis is true is less than the false discovery rate. False discovery rate can fluctuate. In some embodiments, the false discovery rate is 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%. , 14%, 15%, 16%, 17%, 18%, 19%, 20%, or generally such values, or a number or range between any two of these values. In some embodiments, the false discovery rate is at least, or at most, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%. , 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%. The first statistical analysis may be a plurality of binomial tests.

At block 528, the Poisson model can be used to threshold the target's molecular label to determine the true and false molecular label associated with the target in the sequencing data. For example, a Poisson model can be applied to sequencing reads to identify "potentially true" molecular labels from an artificial object.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data will be the molecular label of the target if the target has a complete sequencing quality status. Can be regulated by the step of thresholding to determine the true and false molecular labels associated with the target in the sequencing data. The step of thresholding the target molecular label may include performing a statistical analysis on the target molecular label.

In some embodiments, the steps of performing a statistical analysis are as follows: fitting the distribution of target molecular labels and their number of occurrences to two Poisson distributions; using the two Poisson distributions, a true molecular label. A step of determining the number n; a step of removing the fake molecular label from the sequencing data, wherein the fake molecular label is a lower number of occurrences than the nth richest molecular label. Also, a true molecular label comprises a molecular label having a number of occurrences equal to or greater than the number of occurrences of the nth most abundant molecular label. The two Poisson distributions may include a first Poisson distribution corresponding to a true molecular label and a second Poisson distribution corresponding to a false molecular label.

At block 532, the number of targets can be estimated to generate output after correcting or adjusting the sequencing data using multiple binomial tests or two Poisson distributions. Embodiment 500 ends at the end point block 536.

Saturated Sequencing Quality Status The method disclosed herein is due to the large uncertainty in estimating the molecular label count, if the target has a saturated sequencing quality status, the number of targets during sequencing live. It may not be possible to provide an estimate. With reference to FIG. 5, in some embodiments, in decision state 520, if the sequencing status is not a complete sequencing status, embodiment 500 proceeds to decision state 540. In decision state 540, if the target has a saturated sequencing status, embodiment 500 proceeds to endpoint block 536. For saturated sequencing status, the number of targets may not be determined due to the large uncertainty in estimating the molecular label count.

Incomplete Sequencing Quality Status The methods disclosed herein can provide an estimate of the number of targets in a sequencing library if the target has an incomplete sequencing quality status. When the target in the sequencing library has an incomplete sequencing quality status, the noisy target, eg, the noisy gene, can be eliminated. A target can be noisy if its amplification rate (mean read per molecular label) is similar to the amplification rate of errors derived from fully sequenced genes in the same library containing the target. .. Zero for sequencing reads of targets with incomplete sequencing quality status to extrapolate an estimate of the number of probability barcodes containing targets with identifiable molecular labels present in the library. A cutting Poisson model can be applied.

Embodiment 500 estimates the number of targets in the sequencing library if some of the true probability barcodes used to label the starting targets are not observed due to improper sequencing depth. A value can be provided. In decision state 540, if the target does not have a saturated sequencing status, the target has an incomplete sequencing status and embodiment 500 proceeds to block 544 for a noisy target, eg, a noisy gene. Remove.

If the target's dispersal index is> 4 and the target's maximum sequencing read is> 18, Poisson modeling is used to set a threshold for distinguishing between true and error barcodes. Even if it is acquired, it is still possible to provide a suitable estimated value. If the sequencing data show mild overspraying, eg 1.5 <spreading index ≤4, and the maximum sequencing read for the target is ≤18, then a Poisson model is used to obtain a threshold. And may underestimate the true molecular labeling count. The reason for the underestimation may be that the molecular label with the low read may be a mixture of a true molecular label and a false molecular label. As a result, these true molecular labels with low sequencing leads are forced to enter the Poisson model of error, and the Poisson model of true molecular labels may have fewer molecular labels than they should. For example, an ad hoc method can be used that uses the molecular label count after the low molecular label count, such as 1, has been removed. If the dispersal index is close to 1, eg 0.9-1.5, the observed molecular label counts can be generated with reasonable estimates. If the dispersal index is 0.1-0.9, a zero-cut Poisson model characterized by an under-dispersed Poisson model can produce reasonable estimates; if there are errors in the sequencing data. , This model can tend to be overestimated.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data is such that the quality status of the target in the sequencing data is incomplete sequencing quality status. It can be regulated by a step of determining whether the target is noisy in the sequencing data; and a step of removing the noisy target from the sequencing data. The target can be noisy if the number of molecular labels generated by the noisy target is less than or equal to the incomplete sequencing noisy target threshold. Incomplete sequencing noisy gene thresholds can fluctuate. In some embodiments, the incomplete sequencing noisy target threshold may be equal to the median or average incidence of molecular labels for multiple targets with complete sequencing quality status. In some embodiments, the incomplete sequencing noisy gene threshold is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or roughly such, or any of these values. It can be a number or range between the two. In some embodiments, the incomplete sequencing noisy gene threshold can be at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10.

At block 548, to extrapolate an estimate of the number of probability barcodes containing targets with identifiable molecular labels present in the library, to sequencing reads of targets with incomplete sequencing quality status. In contrast, the zero-cut Poisson model is applied. In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data is such that the quality status of the target in the resulting sequencing data is incompletely sequenced quality status. In some cases, it is regulated by a step of determining whether the target is noisy in the sequencing data; and a step of removing the noisy target.

In some embodiments, the number of molecular labels with identifiable sequences associated with the target in the sequencing data is such that the quality status of the target in the sequencing data is incomplete sequencing quality status. It can be adjusted by the step of thresholding the target molecular label to determine the true molecular label and the false molecular label in the sequencing data. The step of thresholding the target molecular label may include performing a statistical analysis on the molecular label. The step of performing a statistical analysis on a molecular label may include a step of determining the number n of true molecular labels using a zero-cut Poisson model; and a step of removing false molecular labels from the sequencing data.

In some embodiments, the sham molecular label may comprise a molecular label having a lower incidence than the nth most abundant molecular label occurrence . A true molecular label may include a molecular label having an incidence greater than or equal to the nth abundant molecular label occurrence .

Sequencing Data Errors The methods disclosed herein can be used to identify and / or correct sequencing data errors, eg, errors that occur in the method of counting one or more target nucleic acids. In some embodiments, the error may include or be a deletion of one or more nucleotides, a substitution of one or more nucleotides, an addition of one or more nucleotides, or any combination thereof. Errors can be present in molecular labels (ML), sample labels (SL), and other labels on probability barcodes. In some embodiments, the sequencing data error may include, or may be, a PCR induction error, a sequencing induction error, a reverse transcriptase (RT) primer contamination error, or any combination thereof. PCR induction errors can include or be the result of PCR amplification errors, PCR amplification bias, inadequate PCR amplification, or any combination thereof. Sequencing implementation errors may include or be the result of inaccurate base calling, inadequate sequencing, or any combination thereof. The RT primer contamination error can be an error caused by the reverse transcription primer that has entered the PCR.

As used herein, the term "coverage" or "sequencing depth" can mean the number of barcoded target reads with a particular ML and a particular SL in the sequencing data. For example, a bar coded target can be sequenced multiple times. Therefore, a barcoded target with a specific ML and SL can be observed multiple times. As another example, the cell may contain multiple copies of the target (eg, multiple copies of the mRNA molecule of a gene). Multiple copies of these targets can be barcoded. After PCR amplification (eg, block 28 in the figure), there can be multiple copies of the barcoded target with a particular ML and SL. Upon sequencing, some or all of multiple copies of a barcoded target with a particular ML and SL may be sequenced. The number of barcoded target reads with the same ML and SL observed in the sequencing data is sometimes referred to as "coverage" or "sequencing depth".

In some embodiments, sequencing data errors can be identified and / or corrected. For example, a copy of a target from a cell can be barcoded with a different ML and the same SL. A bar coded target with ML can have multiple reads in the sequencing data. Barcoded targets with different MLs may have only a small number of reads (eg, one lead). Earlier barcoded targets may be more likely to have true ML (or real or signal ML) than later barcoded targets. Later barcoded targets may contain error ML (or false or noise ML). This may be because the two MLs can be expected to have similar coverage or sequencing depth. Barcoded targets after containing only a few reads can be artifacts or errors that occur during sequencing or PCR.

As another example, probability barcodes that enter PCR can cause RT primer contamination errors. In some embodiments, after reverse transcription of the mRNA molecule into the cDNA molecule (eg, Figure 24), probability barcodes that are not integrated into the cDNA molecule can be removed, for example, by purification bead purification. Removal methods, such as Aple bead purification, may not completely remove probability barcodes that are not extended by reverse transcription incorporated into the probability barcoded cDNA molecule. For example, 15%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2% of probability barcodes that are not extended by reverse transcription incorporated into a cDNA molecule with probability barcodes. The range between 1%, 0.5%, 0.1%, or any two of these values may not be removed by Ample bead purification. These unremoved probability barcodes can cause sequencing data errors during amplification of cDNA molecules (eg, block 28 in the figure). Probability barcodes can be very similar between samples. For example, the sample markers for probability barcodes can be the same for the same sample. Therefore, these unremoved probability barcodes may hybridize to other nucleic acid molecules from the same sample (eg, the SL region of the mRNA molecule with probability barcode) during PCR, thus crossing the PCR. As a result, a sequencing data error called an SL error can occur.

True ML, error ML, and SL error can have an identifiable distribution. FIG. 4 is a schematic diagram showing a non-limiting exemplary distribution of molecular labeling errors, sample labeling errors, and true molecular labeling signals. As shown in FIG. 4, it is believed that error MLs tend to have lower ML coverage because error MLs can be due to PCR or sequencing errors. For example, error ML can be due to the majority of sequencing errors and some of the PCR errors. It is believed that SL errors tend to have lower ML coverage, as SL errors can be largely due to the probability barcode of entering the PCR.

Correction of PCR and Sequencing Errors Based on Directional Accessibility This specification discloses methods of correcting PCR or sequencing errors. In some embodiments, the method comprises (a) receiving sequencing data for a target with a probability barcode. A target with a probability barcode can be obtained by a process of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a plurality of targets with a probability barcode, and here, a plurality of targets have a probability barcode. Each of the probability barcodes contains a molecular label. In some embodiments, the method is (b) for one or more of a plurality of targets: (i) counting the number of molecular labels having identifiable sequences associated with the targets in the sequencing data. And; (ii) using directional proximity to identify clusters of target molecular labels; and; (iii) using clusters of target molecular labels identified in (ii), received in (b). The step of collapsing the sequencing data; (iv) the step of estimating the number of targets, the estimated number of targets was counted in (i) after folding the sequencing data in (ii). Includes a step that correlates with the number of molecular labels having an identifiable sequence associated with the target in the sequencing data. Multiple targets include targets for the entire transcriptome of the cell. In some embodiments, the method further comprises (c) using multiple probability barcodes to attach probability barcodes to the plurality of targets to generate multiple probability barcoded targets; d) Includes a step of sequencing a target with a probability barcode to generate sequencing data for the received target with a probability barcode.

FIG. 7 is a flow chart illustrating a non-limiting exemplary embodiment 700 that corrects PCR or sequencing errors using molecular labeling based on directional proximity. The process of correcting PCR or sequencing errors using directional proximity-based molecular labeling is sometimes referred to as recursive replacement error correction (RSEC). The method 700 starts at block 704 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, the method 700 further comprises the step of assigning a probability barcode to the plurality of targets to generate a plurality of barcoded targets using the plurality of probability barcodes, wherein the method 700 comprises a plurality of. Each of the probability barcodes of is contained a molecular label. In some embodiments, the method 700 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 708, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At block 712, directional proximity can be used to identify clusters of targeted molecular labels. The target molecular labels within the cluster can be located within a predetermined directional proximity threshold of each other. The directional proximity threshold can fluctuate. In some embodiments, the predetermined directional proximity threshold can be one or two Hamming distances, or generally at least, or at most, such distances.

In some embodiments, the target molecular label within the cluster may include one or more parent molecule labels and one or more child molecule labels of one or more parent molecule labels. The number of occurrences of the parent molecule label may be equal to or higher than the predetermined number of occurrences of directional proximity. In some embodiments, the predetermined directional proximity occurrence number threshold can be 2 × ( number of child molecule labeling occurrences) -1 , or, at least, or at most, such a value. In some embodiments, the predetermined directional proximity occurrence number threshold is 1.5 times, 2 times, 3 times, 4 times, 5 times, 6 times, 7 times, 8 times, the number of occurrences of child molecule labels. It can be 9x, 10x, or roughly such a value, or a number or range between any two of these values. In some embodiments, the predetermined directional proximity occurrence number threshold is at least or at most 1.5 times, 2 times, 3 times, 4 times, 5 times, 6 times, 7 times the number of occurrences of child molecule labels. , 8 times, 9 times, 10 times.

At block 720, the sequencing data is collapsed using a cluster of target molecular labels. The step of collapsing the sequencing data may include a step of assigning the number of occurrences of the child molecule label to the parent molecule label. At block 732, after collapsing the sequencing data, the number of targets can be estimated to generate output. Method 700 ends at block 736.

In some embodiments, the method further comprises the step of determining the sequencing depth of the target. The step of estimating the number of targets includes adjusting the sequencing data counted in (i) when the sequencing depth of the targets exceeds a predetermined sequencing depth threshold. The predetermined sequencing depth threshold may be between 15 and 20. The step of adjusting the sequencing data counted in (i) is to threshold the molecular label of the target and the true molecular label and false molecule associated with the target in the sequencing data obtained in (b). Includes the step of determining the label. The step of thresholding the target molecular label includes a step of performing statistical analysis on the target molecular label. The steps to perform statistical analysis are to apply the distribution of target molecular labels and the number of their occurrences to two distributions, such as two negative binomial distributions; true molecular labeling using two negative binomial distributions. A step of determining the number n of the above; a step of removing the fake molecular label from the sequencing data obtained in (b), wherein the fake molecular label is the nth richest molecular label. A true molecular label comprises a molecular label having a number of occurrences lower than the number of occurrences, and a true molecular label comprises a molecular label having a number of occurrences greater than or equal to the number of occurrences of the nth most abundant molecular label.

Correction of PCR and Sequencing Errors Based on Directional Accessibility and Second Derivative Functions herein discloses a method for determining the number of targets. In some embodiments, one method is (a) using a plurality of probabilistic barcodes to attach probabilistic barcodes to a plurality of targets to generate a plurality of probabilistic barcoded targets. Each of the probabilistic barcodes of the above includes a molecular label; (b) a step of acquiring sequencing data of a target with a probabilistic barcode; (c) for one or more of a plurality of targets: (i) sequencing data. A step of counting the number of molecular labels having an identifiable sequence associated with a target in; (iii) a step of identifying clusters of target molecular labels using directional proximity; (iii) (ii). ) Using the cluster of molecular labels of the targets identified in), the step of collapsing the sequencing data obtained in (b) and the step of estimating the number of targets; (iv) the estimated number of targets. Consists of folding the sequencing data in (ii) and then correlating with the number of molecular labels having identifiable sequences associated with the target in the sequencing data counted in (i). Multiple targets may include targets for the entire transcriptome of the cell.

In some embodiments, the method comprises the step of determining the sequencing status of the target in the sequencing data. The sequencing status of the target in the sequencing data may include saturated sequencing or may be saturated sequencing. In some embodiments, if the sequencing status of the targets in the sequencing data is saturated sequencing status, the number of targets estimated in (iv) will be in the sequencing data counted in (i). Correlates with the number of molecular labels with identifiable sequences associated with the target of.

In some embodiments, the estimated number of targets correlates with the number of molecular labels having identifiable sequences associated with the targets in the sequencing data counted in (i) after correcting the SL error. do. The steps to correct SL errors are to create a cumulative sum plot of molecular labels with identifiable sequences associated with the target in the sequencing data; and to determine the quadratic derivative of the cumulative sum plot; It comprises the step of determining the ML read depth cutoff based on the minimum value of the quadratic derivative of the cumulative sum plot. In some embodiments, the step of correcting SL errors removes molecular labels with identifiable sequences associated with targets in sequencing data that have a read depth lower than the determined ML read depth cutoff. May include steps.

FIG. 8 is a flow chart illustrating a non-limiting exemplary embodiment 800 that corrects PCR and sequencing errors using molecular labeling based on directional proximity and quadratic derivatives. Method 800 begins at block 804 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, the method 800 further comprises the step of attaching a probability barcode to the plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes. Each of the multiple probability barcodes contains a molecular label. In some embodiments, Method 800 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 808, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the sequencing targets can be counted. At decision block 812, it can be determined whether or not the sequencing data has a saturated sequencing status. Saturation sequencing status can be determined by a target having a number of molecular labels with identifiable sequences above a predetermined saturation threshold. A given saturation threshold can vary for different performances. For example, if the probability barcode has a molecular label of about 6651 with an identifiable sequence, the predetermined saturation threshold can be about 6557. As another example, if the probability barcode has a molecular label of about 65536 with an identifiable sequence, the predetermined saturation threshold can be about 65532.

If the sequencing data does not have a saturated sequencing status at decision block 812, method 800 can proceed to block 816, where the molecular label count can be adjusted based on directional proximity. The target is, for example, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 70000, 80000, 90000, 100,000, or these. Having a number of molecular labels with an identifiable sequence that exceeds the number or range between any two of the above can be considered to have saturated sequencing status. As another example, the target exceeds 50%, 60%, 70%, 80%, 90%, 95%, 99%, or 99.9% of the molecular barcode of the probability barcode having an identifiable sequence. , Or a number of molecular labels having an identifiable sequence that exceeds the number or range between any two of these, can be considered to have saturated sequencing status. In some embodiments, the step of adjusting the molecular count based on directional proximity can be described with reference to FIG. For example, the step of adjusting the molecular count based on a dictionary is the step of identifying a cluster of target molecular labels using directional proximity; and the step of collapsing sequencing data using the cluster of identified target molecular labels. A step and a step of estimating the number of targets can be included, wherein the estimated number of targets is associated with the targets in the counted sequencing data after collapsing the sequencing data. Correlates with the number of molecular labels with identifiable sequences.

At block 820, the quadratic derivative of the cumulative sum plot can be determined. The step of determining the quadratic derivative of the cumulative sum plot may include creating a cumulative sum plot of molecular labels with identifiable sequences associated with the target in the sequencing data.

At block 824, the molecular label can be adjusted based on the ML read depth cutoff. The ML read depth cutoff may be based on the minimum value of the quadratic derivative of the cumulative sum plot (eg, the local minimum or the global minimum). In some embodiments, the step of correcting SL errors removes molecular labels with identifiable sequences associated with targets in sequencing data with read depths lower than the determined ML read depth cutoff. May include steps.

At block 828, the number of targets can be estimated, the sequencing data can be collapsed, and the output after correcting the SL error can be generated. In decision block 812, if the sequencing data has a saturated sequencing status, method 800 can proceed to block 828 to generate output without folding the sequencing data and correcting SL errors. Method 800 ends at block 832.

Correction of PCR and Sequencing Errors Based on Directional Accessibility and Correction of Errors Based on Distribution This specification discloses methods of correcting PCR or sequencing errors. The method can be used to determine the number of targets. In some embodiments, the method comprises (a) receiving sequencing data for a target with a probability barcode. A target with a probability barcode can be obtained by a process of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a plurality of targets with a probability barcode, and here, a plurality of targets have a probability barcode. Each of the probability barcodes contains a molecular label. In some embodiments, the method is (b) for one or more of a plurality of targets: (i) counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data. And; (iii) a step of determining the number of noise molecule labels having an identifiable sequence associated with the target in the sequencing data; (iii) a step of estimating the number of targets, wherein. The estimated number of targets is adjusted according to the number of noise molecular labels determined in (ii), molecular labels with identifiable sequences associated with the targets in the sequencing data counted in (i). Correlates with the number of. In some embodiments, the method comprises the step of determining the sequencing status of the target in the sequencing data. In some embodiments, the method further comprises (c) using multiple probability barcodes to attach probability barcodes to the plurality of targets to generate multiple probability barcoded targets; d) Includes a step of sequencing a target with a probability barcode to generate sequencing data for the received target with a probability barcode.

FIG. 9 is a flow chart illustrating a non-limiting exemplary embodiment 900 that corrects PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction. Method 900 starts at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, the method 900 further comprises the step of attaching a probability barcode to the plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes, wherein the method 900 further comprises a plurality of steps. Each of the probability barcodes of the contains a molecular label. In some embodiments, the method 900 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At decision block 912, the sequencing data can determine whether or not it has a saturated sequencing status. For example, the target is that it exceeds 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000, 20000, 30000, 40000, 50000, 60000, 70000, 80000, 90000, 100,000, or these. Having a number of molecular labels with an identifiable sequence that exceeds the number or range between any two of the above can be considered to have saturated sequencing status. As another example, the target exceeds 50%, 60%, 70%, 80%, 90%, 95%, 99%, or 99.9% of the molecular barcode of the probability barcode having an identifiable sequence. , Or a number of molecular labels having an identifiable sequence that exceeds the number or range between any two of these, can be considered to have saturated sequencing status.

In some embodiments, the saturation sequencing status can be determined by a target having a number of molecular labels with identifiable sequences that exceed a predetermined saturation threshold. The predetermined saturation thresholds may be different in different performances. For example, the predetermined saturation thresholds are 1000, 2000, 3000, 4000, 5000, 6000, 6557, 7000, 8000, 9000, 10000, 20000, 30000, 40000, 50000, 60000, 65532, 70000, 80000, 90000, 100000, Or it can be a number or range between any two of these values, or it can be approximately such a value. As another example, a given saturation threshold is at least or at most 1000, 2000, 3000, 4000, 5000, 6000, 6557, 7000, 8000, 9000, 10000, 20000, 30000, 40000, 50000, 60000, 65532. , 70,000, 80,000, 90,000, or 100,000.

In some embodiments, the saturated sequencing status may depend on the number of molecular labels on the probability barcode having an identifiable sequence. For example, if the probability barcode has a molecular label of about 6651 with an identifiable sequence, the predetermined saturation threshold can be about 6557. As another example, if the probability barcode has a molecular label of about 65536 with an identifiable sequence, the predetermined saturation threshold can be about 65532. In some embodiments, the saturated sequencing status may be independent of the number of molecular labels on the probabilistic barcode having an identifiable sequence.

If the sequencing data is at decision block 912 and does not have a saturated sequencing status, method 900 proceeds to block 916, where the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular count based on directional proximity can be described with reference to FIG. For example, the step of adjusting the molecular count based on a dictionary is the step of identifying a cluster of target molecular labels using directional proximity; and the step of collapsing sequencing data using the cluster of identified target molecular labels. Steps; include estimating the number of targets; where the estimated number of targets is identifiable associated with the target in the counted sequencing data after collapsing the sequencing data. Correlates with the number of molecular labels with sequences.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include or may be undersequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status. For example, a target has a depth (eg, average, minimum, or maximum depth) of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, If it is less than, or generally less than, a number or range between 70, 80, 90, 100, or any two of these values, it can be considered to have an undersequencing status. As another example, the target has a depth of at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, If it is less than 80, 90, or 100, it can be considered to have an undersequencing status.

In some embodiments, the undersequencing status can be determined by a target having a depth less than a predetermined undersequencing threshold (eg, mean, minimum, or maximum depth). The undersequencing threshold may be different for different fulfillments. For example, the undersequencing threshold is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or any of these values. It can be a number or range between any two of the above, or roughly such a value. As another example, the undersequencing threshold is at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80. , 90, or 100.

In some embodiments, the undersequencing status may depend on the number of molecular labels on the probability barcode having an identifiable sequence. For example, the probability bar code has an identifiable sequence, 1000, 2000, 3000, 4000, 5000, 6000, 6651, 7000, 8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 65532, 70000, If you have a number or range between 80,000, 90,000, 100,000, or any two of these values, or approximately such a value, the undersequencing threshold can be 10 (or another number of thresholds). As another example, the probability bar code is at least or at most 1000, 2000, 3000, 4000, 5000, 6000, 6651, 7000, 8000, 9000, 10000, 20000, 30000, 40,000, 50000, 60000, 65532, When including 70,000, 80,000, 90,000, or 100,000, the undersequencing threshold can be 10 (or another number of thresholds). In some embodiments, the saturated sequencing status may be independent of the number of molecular labels on the probabilistic barcode having an identifiable sequence.

At decision block 924, if the sequencing status of the target in the sequencing data is not undersequencing status, method 900 can proceed to block 928 to filter the molecular label count. The step of filtering the molecular label count comprises in determination block 932 determining the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold. Pseudo-point thresholds may differ for different fulfillments. For example, if the probability barcode has a molecular label of about 6651 with an identifiable sequence, the pseudopoint thresholds are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30. , 40, 50, 60, 70, 80, 90, 100, or a number or range between any two of these values, or generally such values. As another example, if the probability barcode has about 6651 molecular labels with identifiable sequences, the pseudopoint thresholds are at least, or at most 1, 2, 3, 4, 5, 6, 7, ,. It can be 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100.

In decision block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequence is performed. Before determining the number of noise molecular labels with identifiable sequences associated with a target in the single data, pseudopoint the number of molecular labels with identifiable sequences associated with the target in the sequencing data. Can be added. Pseudopoints can have different molecular label counts for different performances. For example, the molecular label counts for pseudo points are 0.0001, 0.001, 0.01, 0.1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, It can be a number or range between 40, 50, 60, 70, 80, 90, 100, or any two of these values, or roughly such a value. As another example, the molecular label counts for pseudopoints are at least or at most 0.0001, 0.001, 0.01, 0.1, 1, 2, 3, 4, 5, 6, 7, 8 , 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100. In some embodiments, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudopoint threshold, method 900 can proceed to the block and the sequencing data. Method 944 when the number of molecular labels with identifiable sequences associated with the target in is less than the pseudopoint threshold.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold. Non-unique molecular labels can be removed at block 944 to determine the number of noise molecular labels with identifiable sequences associated with the target in the sequencing data. Non-unique molecular labels may include molecular labels with identifiable sequences associated with targets in sequencing data that are greater than a given reused molecular label threshold. Reuse molecular labeling thresholds may differ in different performances. For example, the reuse molecular label threshold is 100, 200, 300, 400, 500, 600, 650, 700, 900, 1000, 2000 if the probability barcode contains about 6651 molecular labels with identifiable sequences. , Or a number or range between any two of these values, or could be approximately such a value. As another example, the reuse molecular labeling threshold is 100, 200, 300, 400, 500, 600, at least, or at most, if the probability barcode contains about 6651 molecular labels with identifiable sequences. It can be 650, 700, 900, 1000, or 2000.

In some embodiments, the step of removing the non-unique molecular label determines the theoretical number of non-unique molecular labels relative to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data. Includes steps. The step of removing the non-unique molecular label comprises the step of removing the molecular label having a larger number of occurrences than the nth richest molecular label of the molecular label having the identifiable sequence associated with the target in the sequencing data. sell. The number n can be a theoretical number of non-unique molecular labels.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. Distribution-based error correction methods may include determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. The step of determining the number of noise molecular labels may include fitting the two negative binomial distributions to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data. For example, the step of determining the number of noise molecule labels is a signal negative binomial distribution (one of the two negative binomial distributions), a molecule with an identifiable sequence associated with the target in the counted sequencing data. A step of fitting to the number of labels may include, where the signal negative binomial distribution is the number of molecular labels having an identifiable sequence associated with the target in the counted sequencing data, which is the signal molecular label. Corresponds to. The step of determining the number of noise molecular labels is to make the noise negative binomial distribution (the other of the two negative binomial distributions) a molecular label having an identifiable sequence associated with the target in the counted sequencing data. It may include a number fitting step, where the noise negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are noise molecular labels. .. The step of determining the number of noise molecule labels may include a step of determining the number of noise molecule labels using the fitted signal negative binomial distribution and the fitted noise negative binomial distribution.

In some embodiments, using the fitted signal-negative binomial distribution and the fitted noise-negative binomial distribution, the step of determining the number of noise molecule labels is identifiable associated with the target in the sequencing data. For each of the sequences: the identifiable sequence comprises the step of determining the signal probability, which is a signal negative binomial distribution. Then, the noise probability of the identifiable array can determine the noise probability which is the noise negative binomial distribution. Furthermore, if the signal probability is less than the noise probability, the identifiable sequence can be determined to be a noise molecule label. In some embodiments, the step of adjusting the molecular label count at block 944 finds less than two peaks (two peaks are required to determine the signal negative binomial distribution and the noise negative binomial distribution). If so, it may include the step of removing a single ton (eg, a single base substitution).

At block 948, the number of targets can be estimated to produce output after proximity-based error correction and distribution-based error correction. In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 900 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment. For example, the number of determined noise molecule labels may be zero.

In decision block 924, if the sequencing status of the target in the sequencing data is under-sequencing status, method 900 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of determined noise molecule labels may be zero. Method 900 ends at block 952.

FIG. 10 is a flowchart illustrating a non-limiting exemplary embodiment 1000 of error correction using two negative binomial distributions. The blocks of method 1000 (eg, blocks 904-952) are described with reference to FIG. Briefly, method 1000 begins at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, the method 1000 further comprises the step of attaching a probability barcode to the plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes, wherein the plurality of. Each of the probability barcodes of is contained a molecular label. In some embodiments, Method 1000 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At block 916, the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular label count based on directional proximity can be described with reference to FIG.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include under-sequencing or may be under-sequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status.

At decision block 924, if the target sequencing status in the sequencing data is not under-sequencing status, method 1000 can optionally proceed to decision block 1004. At the determination block 1004, the target sequencing depth can be compared to a predetermined sequencing depth threshold. Sequencing depth thresholds may be different for different fulfillments. For example, the target sequencing depths are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or these. It can be a number or range between any two of the values, or roughly such a value. As another example, the target sequencing depth is at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, It can be 80, 90, or 100.

If the target sequencing depth is greater than the sequencing depth threshold, method 1000 proceeds to block 928. If the target sequencing depth is less than or equal to the sequencing depth threshold, method 1000 proceeds to block 1008. At block 1008, singletons (eg, single base substitutions) can be removed prior to the step of producing the output of block 948.

At block 928, the molecular label count can be filtered. The step of filtering the molecular label count can include, at decision block 912, the step of determining whether or not the sequencing data has a saturated sequencing status. If the sequencing data is in decision block 912 and does not have a saturated sequencing status, method 1000 can proceed to decision block 932. At the determination block 932, the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold, can be determined.

In decision block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequence is performed. Before determining the number of noise molecular labels with identifiable sequences associated with a target in the single data, pseudopoint the number of molecular labels with identifiable sequences associated with the target in the sequencing data. Can be added. In some embodiments, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudopoint threshold, method 900 can proceed to the block and the sequencing data. Method 944 when the number of molecular labels with identifiable sequences associated with the target in is less than the pseudopoint threshold.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold. Non-unique molecular labels can be removed at block 944 to determine the number of noise molecular labels with identifiable sequences associated with the target in the sequencing data. Non-unique molecular labels may include molecular labels with identifiable sequences associated with targets in sequencing data that are greater than a given reused molecular label threshold.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. Distribution-based error correction methods may include determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. The step of determining the number of noise molecular labels has two negative binomial distributions, i.e., a signal negative binomial distribution and a noise negative binomial distribution, with an identifiable sequence associated with the target in the sequencing data. It may include a step that fits into the number of molecular labels. The signal negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are signal molecular labels. The noise-negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are noise molecular labels. The step of determining the number of noise molecule labels can include a step of determining the number of noise molecule labels using the fitted signal negative binomial distribution and the fitted noise negative binomial distribution.

At block 948, the number of targets can be estimated to produce output after proximity-based error correction and distribution-based error correction. In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 1000 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment.

In decision block 924, if the target sequencing status in the sequencing data is under-sequencing status, method 1000 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of determined noise molecule labels may be zero. Method 1000 ends at block 952.

PCR and Sequencing Error Correction Based on Directional Accessibility, Error Correction Based on Distribution, and Subsampling FIG. 11 shows a non-limiting exemplary embodiment 1100 of error correction using two negative binomial distributions. It is a flowchart. The blocks of method 1100 (eg, blocks 904-952) are described with reference to FIG. Briefly, method 1100 begins at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, method 1100 further comprises the step of assigning a plurality of probabilistic barcodes to a plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes. Each of the probability barcodes of is contained a molecular label. In some embodiments, method 1100 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At decision block 912, it can be determined whether or not the sequencing data has a saturated sequencing status. If the sequencing data is at decision block 912 and does not have a saturated sequencing status, method 1100 proceeds to block 916, where the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular count based on directional proximity can be described with reference to FIG.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include or may be undersequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status.

At decision block 924, if the sequencing status of the target in the sequencing data is not under-sequencing status, method 1100 can optionally proceed to decision block 1104. At decision block 1104, it can be determined whether the sequencing status of the target in the sequencing data is excessive sequencing data. For example, a target may have a depth (eg, average, minimum, or maximum depth) of 50, 100, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000, or any of these values. If it is greater than the number or range between the two, or generally greater than such a value, it can be considered to have an oversequencing status or highly expressed target. As another example, a target has an undersequencing status if its depth is at least, or at most, greater than 50, 100, 200, 250, 300, 400, 500, 600, 700, 800, 900, or 1000. Can be considered to have.

In some embodiments, the oversequencing status or highly expressed target can be determined by a target having a depth greater than a predetermined oversequencing threshold (eg, mean, minimum, or maximum depth). Excessive sequencing thresholds can be different for different performances. For example, the excess sequencing threshold is a number or range between 50, 100, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000, or any two of these values, or generally such. Can be a value. As another example, the excess sequencing threshold can be at least, or at most, 50, 100, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000.

In some embodiments, the oversequencing status may depend on the number of molecular labels on the probabilistic barcode having an identifiable sequence. For example, if the probability barcode contains about 6651 molecular labels with identifiable sequences, the excess sequencing thresholds are 50, 100, 200, 250, 300, 400, 500, 600, 700, 800, 900, It can be 1000, or a number or range between any two of these values, or roughly such a value. As another example, if the probability barcode contains about 6651 molecular labels with identifiable sequences, the excess sequencing threshold is at least, or at most, 50, 100, 200, 250, 300, 400, 500. , 600, 700, 800, 900, 1000. In some embodiments, the undersequencing status may not depend on the number of molecular labels on the probability barcode having an identifiable sequence.

In decision block 1104, if the target has an excessive sequencing status, method 1100 proceeds to block 1108. At block 1108, target ML coverage can be reduced, for example, by subsampling the ML coverage of all targets. For example, the number of molecular labels with identifiable sequences associated with a target in sequencing data can be subsampled to an approximation of a given excess sequencing threshold for all targets (eg, 10). Method 1100 proceeds from block 1108 to block 928.

In determination block 1104, if the target does not have excessive sequencing status, method 1100 proceeds to block 928 to filter the molecular label count. The step of filtering the molecular label count may include in determination block 932 determining the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold.

In decision block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequence is performed. Before determining the number of noise molecular labels with identifiable sequences associated with a target in the single data, pseudopoint the number of molecular labels with identifiable sequences associated with the target in the sequencing data. Can be added. In some embodiments, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudopoint threshold, method 900 can proceed to the block and the sequencing data. Method 944 when the number of molecular labels with identifiable sequences associated with the target in is less than the pseudopoint threshold.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold. Non-unique molecular labels can be removed at block 944 to determine the number of noise molecular labels with identifiable sequences associated with the target in the sequencing data. Non-unique molecular labels may include molecular labels with identifiable sequences associated with targets in sequencing data that are greater than a given reused molecular label threshold.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. Distribution-based error correction methods may include determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. The step of determining the number of noise molecular labels has two negative binomial distributions, i.e., a signal negative binomial distribution and a noise negative binomial distribution, with an identifiable sequence associated with the target in the sequencing data. It may include a step that fits into the number of molecular labels. The signal negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are signal molecular labels. The noise-negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are noise molecular labels. The step of determining the number of noise molecule labels includes a step of determining the number of noise molecule labels using the fitted signal negative binomial distribution and the fitted noise negative binomial distribution.

After adjusting the molecular label count with distribution-based error correction in block 944, method 1100 optionally proceeds to block 1112. At block 1112, the adjusted molecular labeling count from block 944 can be combined with the molecular labeling count determined at block 916 and adjusted based on directional proximity. For example, non-unique molecular labels are removed in block 940 and are not used for distribution fitting in block 944. However, these molecular labels are still present in the molecular label count determined in block 916 and adjusted based on directional proximity. Thus, the adjusted molecular labeling count from block 944 and the adjusted molecular labeling count at block 944 can be combined to generate an output at block 948.

In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 1100 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment. In decision block 924, if the sequencing status of the target in the sequencing data is undersequencing status, method 1100 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of noise molecule labels determined can be zero. Method 1100 may end, for example, at block 952.

FIG. 12 is a flowchart illustrating a non-limiting exemplary embodiment 1200 of error correction using two negative binomial distributions. The blocks of method 1200 (eg, blocks 904-952 and blocks 1104) are described with reference to FIGS. 9 and 11. Briefly, method 1200 begins at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, method 1200 further comprises the step of assigning a plurality of probabilistic barcodes to a plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes. Each of the probability barcodes of is contained a molecular label. In some embodiments, method 1200 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At decision block 912, it can be determined whether or not the sequencing data has a saturated sequencing status. If the sequencing data is at decision block 912 and does not have a saturated sequencing status, method 1200 proceeds to block 916, where the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular count based on directional proximity can be described with reference to FIG.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include or may be undersequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status.

At decision block 924, if the sequencing status of the target in the sequencing data is not under-sequencing status, method 1200 can optionally proceed to decision block 1104. At decision block 1104, it can be determined whether the sequencing status of the target in the sequencing data is excessive sequencing data.

In determination block 1104, if the target has an oversequencing status or the target is a highly expressed target, method 1200 proceeds to block 1208, optionally. At block 1208, the ML coverage of the target can be reduced, for example, by subsampling the ML coverage for each target. For example, the number of molecular labels with identifiable sequences associated with a target in the sequencing data can be subsampled per target to an approximation of a predetermined excess sequencing threshold. Method 1200 proceeds from block 1208 to block 928.

In determination block 1104, if the target does not have excessive sequencing status, method 1200 proceeds to block 928 to filter the molecular label count. The step of filtering the molecular label count may include in determination block 932 determining the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold.

At block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequencing. Arbitrary pseudopoints in the number of molecular labels with identifiable sequences associated with the target in the sequencing data before determining the number of noise molecule labels with identifiable sequences associated with the target in the data. It can be added by selection. In some embodiments, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudopoint threshold, method 900 can proceed to the block and the sequencing data. Method 944 when the number of molecular labels with identifiable sequences associated with the target in is less than the pseudopoint threshold.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold. Non-unique molecular labels can be removed at block 944 to determine the number of noise molecular labels with identifiable sequences associated with the target in the sequencing data. Non-unique molecular labels may include molecular labels with identifiable sequences associated with targets in sequencing data that are greater than a given reused molecular label threshold.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. Distribution-based error correction methods may include determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. The step of determining the number of noise molecular labels has two negative binomial distributions, i.e., a signal negative binomial distribution and a noise negative binomial distribution, with an identifiable sequence associated with the target in the sequencing data. It may include a step that fits into the number of molecular labels. The signal negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are signal molecular labels. The noise-negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are noise molecular labels. The step of determining the number of noise molecule labels includes the step of determining the number of noise molecule labels using the fitted signal-negative binomial distribution and the fitted noise-negative binomial distribution.

After adjusting the molecular label count with distribution-based error correction in block 944, method 1200 proceeds to block 1112, optionally. At block 1112, the adjusted molecular labeling count from block 944 can be combined with the molecular labeling count determined at block 916 and adjusted based on directional proximity. For example, non-unique molecular labels are removed in block 940 and are not used for distribution fitting in block 944. However, these molecular labels are still present in the molecular label count determined in block 916 and adjusted based on directional proximity. Thus, the adjusted molecular labeling count from block 944 and the adjusted molecular labeling count at block 944 can be combined to generate an output at block 948.

In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 1200 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment. In decision block 924, if the sequencing status of the target in the sequencing data is undersequencing status, method 1200 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of noise molecule labels determined can be zero. Method 1200 ends at block 952.

PCR and Sequencing Error Correction Based on Directional Proximity and Distribution-Based Error Correction Using Initial Parameter Estimate for Distribution Fit Figure 13 is PCR based on recursive recursive replacement error correction and distribution-based error correction. It is a flowchart which shows the non-limiting exemplary embodiment 13 of the correction of a sequencing error. The blocks of the method 1300 (eg, blocks 904-952) are described with reference to FIG. Briefly, method 1300 begins at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, the method 1300 further comprises the step of attaching a probability barcode to the plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes, wherein the plurality of. Each of the probability barcodes of is contained a molecular label. In some embodiments, method 1300 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At decision block 912, it can be determined whether or not the sequencing data has a saturated sequencing status. If the sequencing data is at decision block 912 and does not have a saturated sequencing status, method 1300 proceeds to block 916, where the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular label count based on directional proximity can be described with reference to FIG.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include or may be undersequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status.

At decision block 924, if the target sequencing status in the sequencing data is not under-sequencing status, method 1300 can proceed to block 928 to filter the molecular label count. The step of filtering the molecular label count may include in determination block 932 determining the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold.

In decision block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequence is performed. Before determining the number of noise molecular labels with identifiable sequences associated with a target in the single data, pseudopoint the number of molecular labels with identifiable sequences associated with the target in the sequencing data. Can be added. In some embodiments, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudopoint threshold, method 900 can proceed to the block and the sequencing data. Method 944 when the number of molecular labels with identifiable sequences associated with the target in is less than the pseudopoint threshold.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold.

The initial parameters of the two negative binomial distributions can be optionally estimated in block 1304 before adjusting the molecular label count in block 944. The initial parameters of the two negative binomial distributions may be different in different implementations. In some embodiments, the average and dispersal degree of each of the two negative binomial distributions can be 1. In some embodiments, the average and dispersal of the two negative binomial distributions can be estimated to be the average and dispersal of a non-empty subset of the filtered molecular label counts from block 928. For example, the subset can be a 25% to 75% quantile of the filtered molecular label count from block 928. The upper or lower bounds of these quantiles may be different in different performances. In some embodiments, the upper or lower bounds are 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 20%, 30%, 40%. , 50%, 70%, 80%, 90%, 99%, or any number or range between two of these values, or an approximation thereof. In some embodiments, the upper or lower bound is at least or at most 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 20%. , 30%, 40%, 50%, 70%, 80%, 90%, 99%, or 100%.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. Distribution-based error correction methods may include determining the number of noise molecule labels with identifiable sequences associated with the target in the sequencing data. The step of determining the number of noise molecular labels has two negative binomial distributions, i.e., a signal negative binomial distribution and a noise negative binomial distribution, with an identifiable sequence associated with the target in the sequencing data. It may include a step that fits into the number of molecular labels. The signal negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are signal molecular labels. The noise-negative binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the target in the counted sequencing data, which are noise molecular labels. The step of determining the number of noise molecule labels can include a step of determining the number of noise molecule labels using the fitted signal negative binomial distribution and the fitted noise negative binomial distribution.

At block 948, the number of targets can be estimated to produce output after proximity-based error correction and distribution-based error correction. In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 1300 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment.

In decision block 924, if the sequencing status of the target in the sequencing data is under-sequencing status, method 1300 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of determined noise molecule labels may be zero. Method 1300 ends, for example, at block 952.

FIG. 14 is a non-limiting exemplary implementation of PCR and sequencing error correction based on recursive replacement error correction and distribution-based error correction by using the second highest molecular label for initial parameter estimates. It is a flowchart which shows a form. The blocks of the method 1400 (eg, blocks 904-952) are described with reference to FIG. Briefly, method 1400 begins at block 904 after receiving sequencing data for targets with multiple probability barcodes. In some embodiments, method 1400 further comprises the step of assigning a plurality of probabilistic barcodes to a plurality of targets to generate a plurality of probabilistic barcoded targets using the plurality of probability barcodes. Each of the probability barcodes of is contained a molecular label. In some embodiments, method 1400 further comprises the step of sequencing a plurality of probabilistic barcoded targets to obtain sequencing data.

At block 908, for one or more of the targets: the number of molecular labels with identifiable sequences associated with the targets in the sequencing data can be counted. At decision block 912, it can be determined whether or not the sequencing data has a saturated sequencing status. If the sequencing data is at decision block 912 and does not have a saturated sequencing status, method 1400 proceeds to block 916, where the molecular label count can be adjusted based on directional proximity. In some embodiments, the step of adjusting the molecular label count based on directional proximity can be described with reference to FIG.

At block 920, the sequencing status of the target in the sequencing data can be determined. The sequencing status of the target in the sequencing data may include or may be undersequencing. At decision block 924, it is possible to determine whether the sequencing status of the target in the sequencing data is under-sequencing status.

At decision block 924, if the target sequencing status in the sequencing data is not under-sequencing status, method 1400 can proceed to block 928 to filter the molecular label count. The step of filtering the molecular label count may include in determination block 932 determining the number of molecular labels having an identifiable sequence associated with the target in the sequencing data, which is less than the pseudo-point threshold.

In decision block 932, if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is less than the pseudo-point threshold, method 900 optionally proceeds to block 936, where sequence is performed. Before determining the number of noise molecular labels with identifiable sequences associated with a target in the single data, pseudopoint the number of molecular labels with identifiable sequences associated with the target in the sequencing data. Can be added.

In decision block 932, non-unique molecular labels can be removed in block 940 if the number of molecular labels with identifiable sequences associated with the target in the sequencing data is greater than or equal to the pseudopoint threshold.

At block 944, a distribution-based error correction method can be used to adjust the molecular label count. The initial parameters for a distribution-based error correction method may be based on the count of molecular labels. For example, one initial parameter (eg, mean and dispersal) of one of the negative binomial distributions (eg, signal negative binomial distribution or noise negative binomial distribution) is the average or count of molecular labels or the number of molecular labels. It may be based. This molecular label may be the second highest count molecular label or any graded (eg, tenth highest count) molecular label. The grading of molecular labels can vary in various practices. In some embodiments, the grading is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or It can be a number or range between any two of these values, or an approximation of them. In some embodiments, the grading is at least, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80. , 90, or 100. The number of molecular labels can vary in different implementations. In some embodiments, the number of molecular labels is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100. , Or a number or range between any two of these values, or an approximation thereof. In some embodiments, the number of molecular labels is at least or at most 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70. , 80, 90, or 100.

At block 948, the number of targets can be estimated to produce output after proximity-based error correction and distribution-based error correction. In decision block 912, if the sequencing status of the target in the sequencing data is saturated sequencing status, method 1400 proceeds to block 948 for molecular labeling based on directional proximity and distribution-based error correction. The output can be generated without adjustment.

In decision block 924, if the sequencing status of the target in the sequencing data is under-sequencing status, method 1400 proceeds to block 948 without adjusting the molecular label based on distribution-based error correction. , Can produce output. For example, the number of determined noise molecule labels may be zero. Method 1400 ends at block 952.

Sequencing In some embodiments, the step of estimating the number of probabilistic bar coded targets is a labeled target, spatial label, molecular label, sample label, cell label, or any product thereof (eg, a labeled amplicon, or). It may include the step of determining the sequence of the labeled cDNA molecule). Amplified targets can be sequenced. The step of sequencing a probabilistic bar coded target or any product thereof is at least part of the sample label, spatial label, cell label, molecular label, at least part of the probabilistic bar coded target, its complementary strand, reverse complement. A step of performing a sequencing reaction may be included to determine the sequence of the chain, or any combination thereof.

Sequencing of targets with probabilistic barcodes (eg, amplified nucleic acids, labeled nucleic acids, cDNA copies of labeled nucleic acids, etc.) can be performed using a variety of sequencing methods and is limited as such methods. Sequencing by hybridization (SBH), sequencing by ligation (SBL), quantitative incremental fluorescent nucleotide addition sequencing (QIFNAS), stepwise ligation and cleavage, fluorescence resonance energy transfer (FRET), molecular beacons, but not. TaqMan Reporter Probe Digestion, Pyro Sequencing, Fluorescent In situ Sequencing (FISSEC), FISSEQ Beads, Wobble Sequencing, Multiple Sequencing, POLONY Sequencing; Nanogrid Rolling Circle Sequencing, A single template molecule OLA using an allogeneic-specific oligoligation assay (eg, oligoligation (OLA), ligated linear probe and rolling circle amplification (RCA) readout, ligated tablet probe, or ligated. Examples include a single template molecule OLA using a circular tablet probe and rolling circle amplification (RCA).

In some embodiments, the steps of sequencing a stochastic barcode target or any product thereof are paired-end sequencing, nanopore sequencing, high throughput sequencing, shotgun sequencing, dieterminator sequencing, multiple primer DNA. Includes sequencing, primer walking, including Sanger dideoxy sequencing, Maxam Gilbert sequencing, pyrosequencing, true single molecule sequencing, or any combination thereof. Alternatively, the sequence of the probabilistic bar coded target or any product thereof can be determined by electron microscopy or a chemical-sensitive field effect transistor (chemFET) array.

Roche 454, Illumina Solexa, ABI-SOLiD, ION
High-throughput sequencing methods such as circular array sequencing using platforms such as Torrent, Complete Genomics, Pacific Biosciences, Helicos, or Polonator platforms can also be used. In some embodiments, the sequencing may include MiSeq sequencing. In some embodiments, the sequencing may include HiSeq sequencing.

Targets with probability barcodes may contain nucleic acids that occupy about 0.01% to about 100% of the genes in the genome of the organism. For example, by capturing a gene containing a complementary sequence in a sample using a target complementary region containing multiple multimers, approximately 0.01% of the gene in the genome of the organism to about 100% of the gene in the genome of the organism. Can be sequenced. In some embodiments, the probabilistic bar coded target comprises nucleic acids that make up about 0.01% of the transcriptome transcript of the organism to about 100% of the transcriptome transcriptome of the organism. For example, by capturing mRNA from a sample using a target complementary region containing a poly (T) tail, approximately 0.501% of the transcriptome transcript of the organism to the transcriptome transcriptome of the organism. About 100% can be sequenced.

The steps of determining the sequence of spatial and molecular labels for multiple probability barcodes are 0.00001%, 0.0001%, 0.001%, 0.01%, 0.1%, of the multiple probability barcodes. 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80% , 90%, 99%, 100%, or a number or range between any two of these values may be sequenced. The step of determining the sequence of a plurality of probabilistic barcode labels, such as sample labels, spatial labels, and molecular labels, is 1, 10, 20, 30, 40, 50, 60, 70, 80 of the plurality of probabilistic barcodes. , 90, 100, 10 ³ , 10 ⁴ , 10 ⁵ , 10 ⁶ , 10 ⁷ , 10 ⁸ , 10 ⁹ , 10 ¹⁰ , 10 ¹¹ , 10 ¹² , 10 ¹³ , 10 ¹⁴ , 10 ¹⁵ , 10 ¹⁶ , 10 ¹⁷ , It may include the steps of sequencing a number or range between 10 ¹⁸ , 10 ¹⁹ , 10 ²⁰ or any two of these values. The steps of sequencing some or all of multiple probability barcodes are about, at least, or at most 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400. , 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000, or a number or range of nucleotides or bases between any two of these values. It may include the step of producing a read length sequence.

The sequencing step may include sequencing at least or at least about 10, 20, 30, 40, 50, 60, 70, 80, 90, 100 or more nucleotides or base pairs of a probabilistic barcoded target. For example, the sequencing step involves performing polymerase chain reaction (PCR) amplification against multiple probabilistic barcoded targets to generate sequencing data with read lengths of 50, 75, or 100 or more nucleotides. Can include. The sequencing step may include sequencing at least or at least about 200, 300, 400, 500, 600, 700, 800, 900, 1,000 or more nucleotides or base pairs of a probabilistic barcoded target. Sequencing steps include sequencing at least or at least about 1,500, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, or 10000 or more nucleotides or base pairs of probabilistic barcoded targets. sell.

The sequencing step may include at least about 200, 300, 400, 500, 600, 700, 800, 900, 1,000 or more sequencing leads per run. In some embodiments, the sequencing step may include at least or at least about 1,500, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, or 10000 or more sequencing leads per run. The sequencing step may include approximately 1,600,000,000 or less sequencing leads per run. The sequencing step may include up to about 200,000,000 leads per run.

Samples In some embodiments, one or more samples may contain multiple labels. One sample may contain one or more cells, or nucleic acids from one or more cells. One sample may be a single cell or a nucleic acid derived from one cell. The one or more cells may be of one or more cell types. At least one of one or more cell types is brain cells, heart cells, cancer cells, circulating tumor cells, organ cells, epithelial cells, metastatic cells, benign cells, primary cells, circulating cells, or any combination thereof. Is.

Samples for use in the methods of the present disclosure may contain one or more cells. Sample means one or more cells. In some embodiments, the plurality of cells may comprise one or more cells. At least one of one or more cell types is brain cells, heart cells, cancer cells, circulating tumor cells, organ cells, epithelial cells, metastatic cells, benign cells, primary cells, circulating cells, or any combination thereof. May be. In some embodiments, the cell is a cancer cell resected from a cancerous tissue such as breast cancer, lung cancer, colon cancer, prostate cancer, ovarian cancer, pancreatic cancer, brain cancer, melanoma and non-melanoma skin cancer. .. In some cases, cells are derived from cancer but are taken from body fluids (eg, circulating tumor cells). Non-limiting examples of cancers include adenomas, adenomas, squamous cell carcinomas, basal cell carcinomas, small cell carcinomas, large undifferentiated carcinomas, chondrosarcomas, and fibrosarcoma. Samples can include tissue, cell monolayers, fixatives, tissue pieces, or any combination thereof. The sample may include a biological sample, a clinical sample, an environmental sample, a biological fluid, a tissue, or cells from a subject. Samples can be obtained from humans, mammals, dogs, rats, mice, fish, flies, helminths, plants, fungi, bacteria, viruses, vertebrates, or non-vertebrates.

In some embodiments, the cell is a cell that is infected with a virus and contains a viral oligonucleotide. In some embodiments, the viral infection can be caused by a virus such as a single-stranded (+ strand or "sense") DNA virus (eg, parvovirus), or a double-stranded RNA virus (eg, retrovirus). .. In some embodiments, the cell is a bacterium. These may include either Gram-positive or Gram-negative bacteria. In some embodiments, the cell is a fungus. In some embodiments, the cell is a protozoa or other parasite.

As used herein, the term "cell" can mean one or more cells. In some embodiments, the cell is a normal cell, eg, a human cell at different developmental stages, or a human cell derived from different organs or histological types. In some embodiments, it is a non-human cell, eg, another type of mammalian cell (eg, mouse, rat, pig, dog, bovine, or horse). In some embodiments, the cell is another type of animal or plant cell. In other embodiments, the cell can be any prokaryotic cell or eukaryotic cell.

As used herein, cells are sorted before associating the cells with beads. For example, cells can be sorted by fluorescence activated cell sorting or magnetically activated cell sorting, or more generally by flow cytometry. Cells can be filtered by size. In some embodiments, the retainate contains cells associated with the beads. In some embodiments, the flow-through contains cells associated with the beads.

A sample can mean multiple cells. The sample can mean a monolayer of cells. The sample can mean a thin section (eg, tissue flakes). A sample can mean a solid or semi-solid collection of cells that can be placed in a one-dimensional array.

Data Analysis and Display Software Data Analysis and Visualization of Spatial Resolution of Targets The present disclosure provides a method for estimating the number and location of targets using digital counting using probabilistic barcoding and spatial labeling. The data obtained from the methods of the present disclosure can be visualized on a map. Maps of the number and location of targets in the sample can be constructed using the information generated using the methods described herein. Maps can be used to determine the physical position of the target. Maps can be used to identify the location of multiple targets. Multiple targets can be the same species of target, or multiple targets can be multiple different targets. For example, it is possible to build a map of the brain to show the digital counts and positions of multiple targets.

Maps can be generated from a single sample of data. Maps can be constructed using data from multiple samples, thereby generating combinatorial maps. Maps can be constructed with data from dozens, hundreds, and / or thousands of samples. A map composed of multiple samples can show the distribution of digital counts of targets associated with regions common to multiple samples. For example, the replicate assay can be displayed on the same map. At least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 replicates or more may be displayed (eg, overlay) on the same map. At most 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 replicates or more can be displayed (eg, overlay) on the same map. The spatial distribution and number of targets can be represented by various statistics.

By combining data from multiple samples, the positional resolution of the combined map can be increased. The orientation of multiple samples can be registered by common landmarks, where individual position measurements across the sample are at least partially discontinuous. A specific example is to use a microtome to cut a sample on one axis and then cut a second sample along another axis. The combined database will provide a three-dimensional spatial position with a digital count of the target. Multiplexing the above approach will enable high resolution 3D maps of digital count statistics.

In some embodiments of the instrument system, the system will include a computer-readable medium containing code to provide data analysis of the sequence data set generated by performing a single cell probability bar coding assay. Examples of data analysis functions that can be provided by data analysis software are, but are not limited to, (i) sample labels, cells provided by sequencing the probabilistic bar code library generated during the assay. Algorithms for decoding / demultiplexing of labeling, spatial labeling, molecular labeling, and target sequence data, (ii) algorithms for determining read count / gene / cell and unique transcript molecular number / gene / cell, (Iii) Statistical analysis of sequence data, eg, for clustering cells with gene expression data or for predicting confidence intervals in decisions such as transcript molecule number / gene / cell, (iv), eg, principal component analysis. , Hierarchical clustering, k-mean clustering, self-organizing maps, algorithms for identifying rare cell subpopulations, (v) for aligning gene sequence data to known reference sequences. And mutations, polymorphic markers, and sequence alignment capabilities for detecting splice variants, as well as (vi) automatic clustering of molecular labels to compensate for amplification or sequencing errors. In some embodiments, commercial software may be used to perform all or part of the data analysis. For example, Seven Bridges (https://www.sbgenomics.com/) software may be used to edit a table of copies of one or more genes present in each cell in a whole cell collection. In some embodiments, the data analysis software may include options for outputting useful graphed sequencing results, eg, heatmaps showing the number of copies of one or more genes present in each cell of a cell population. .. In some embodiments, the data analysis software has, for example, a copy number of one or more genes present in each cell of a cell population and a type of cell, a type of rare cell type, or a specific disease or condition. By correlating with cells from the subject, further algorithms for extracting biological significance from the sequencing results may be included. In certain embodiments, the data analysis software may further include an algorithm for comparing cell populations across different biological samples.

In some embodiments, all of the data analysis functions may be packaged in a single software package. In some embodiments, the complete set of data analysis capabilities may include a complete set of software packages. In some embodiments, the data analysis software can be a stand-alone package available to the user independent of the assay instrument system. In some embodiments, the software can be web-based as well as allow users to share data.

In some embodiments, all of the data analysis functionality can be packaged in a single software package. In some embodiments, the complete set of data analysis capabilities may include a complete set of software packages. In some embodiments, the data analysis software may be a user-available stand-alone package independent of the assay instrument system. In some embodiments, the software can be web-based and allow users to share data.

System Processors and Networks Generally, computers or processors suitable for use in the equipment system methods of the present disclosure are media that can optionally be connected to server 1509 with fixed media 1512, as shown in FIG. It can be further understood as a logical device that can read instructions from 1511 or network port 1505. The system 1500 may include a CPU 1501, a disk drive 1503, optional input devices such as a keyboard 1515 and a mouse 1516, and an optional monitor 1507, as shown in FIG. Data communication is achievable via a designated communication medium for a server at a local or remote location. The communication medium may include any means of transmitting and receiving data. For example, the communication medium can be a network connection, a wireless connection, or an internet connection. Such a connection can provide communication via the World Wide Web. The data relating to the present disclosure can be transmitted for reception or viewing by a party 1522 over such a network or connection, as shown in FIG.

An exemplary embodiment of the first architectural example of the computer system 1600 shown in FIG. 16 can be used in connection with the embodiments of the present disclosure. As shown in FIG. 16, an example computer system may include a processor 1602 for processing instructions. Examples of processors include, but are not limited to, Intel Xeon ™ processor, AMD Opteron ™ processor, Samsung 32-bit RISC ARM 1176JZ (F) -S v1.0 ™ processor, ARM Cortex-. Examples include A8 Samsung S5PC100 ™ processor, ARM Cortex-A8 Apple A4 ™ processor, Marvel PXA 930 ™ processor, or functionally equivalent processor. Multithreading of execution can be used for parallel processing. In some embodiments, multiple processors, whether a single computer system with cluster connections or a distributed system with network connections that includes multiple computers, mobile phones, or personal digital assistant devices. Alternatively, a processor with multiple cores can be used.

As shown in FIG. 16, the fast cache 1604 can be connected to or installed in the processor 1602 to provide fast memory for recently used or frequently used instructions or data. The processor 1602 can be connected to the north bridge 1606 by the processor bus 1608. The northbridge 1606 is connected to the random access memory (RAM) 1610 by the memory bus 1612 and manages access to the RAM 1610 by the processor 1602. Northbridge 1606 can also be connected to Southbridge 1614 by chipset bus 1616. The south bridge 1614 is eventually connected to the peripheral bus 1618. The peripheral device bus can be, for example, a PCI, PCI-X, PCI Express, or other peripheral device bus. Northbridges and southbridges, often referred to as processor chipsets, manage data transfer between the processor and RAM and peripheral elements on the peripheral bus 1618. In some alternative architectures, the functionality of the northbridge is embedd in the processor instead of using a separate northbridge chip.

In some embodiments, the system 1600 may include an accelerator card 1622 coupled to a peripheral bus 1618. Accelerators may include field programmable gate arrays (FPGAs) or other hardware to accelerate certain processes. For example, accelerators can be used for adaptive data restructuring or to evaluate algebraic expressions used in extended set processing.

Software and data are stored in external storage device 1624 and can be loaded into RAM 1610 or cache 1604 for use by the processor. System 1600 includes an operating system for management system resources. Examples of operating systems are, but are not limited to, Linux®, Windows®, MACOS®, BlackBerry OS®, iOS®, and other functionally equivalent. It includes an operating system as well as application software running on the operating system for managing data storage and optimization according to embodiments of the present invention.

In this example, system 1600 also includes network interface cards (NICs) 1620 and 1621, which provide network interfaces to external storage devices such as network-attached storage (NAS) and other computer systems that can be used for distributed parallel processing. Connected to the provided peripheral bus.

FIG. 17 includes a plurality of computer systems 1702a and 1702b, a plurality of mobile phones and personal digital assistants 1702c, and network-attached storage (NAS) 1704a, and 1704b suitable for use in the methods of the present disclosure. An exemplary diagram of the network 1700 is shown. In embodiments, systems 1712a, 1712b, and 1712c can manage data storage and optimize data access to data stored in network-attached storage (NAS). Mathematical models can be used for the data and can be evaluated using distributed parallel processing computer systems 1712a and 1712b, as well as mobile phones and personal digital assistant systems 1712c. Computer systems 1712a and 1712b, as well as mobile phone and personal mobile information terminal systems 1712c, can also provide parallel processing for adaptive data restructuring of data stored in network-attached storage (NAS) 1714a and 1714b. Is. FIG. 17 is only an example, and a wide variety of other computer architectures and systems can be used in connection with various embodiments of the present invention. For example, a blade server can be used to provide parallel processing. Processor blades can be connected via the backplane to provide parallel processing. The storage device may also be connected to the backplane or may exist as a network attached storage device (NAS) via a separate network interface.

In some embodiments, the processor can hold a separate memory space and can transmit data over the network interface to the backplane or to other connectors for parallel processing by other processors. In other embodiments, some or all of the processors can use the shared virtual address memory space.

An exemplary block diagram of the multiprocessor computer system 1800 shown in FIG. 18 uses a shared virtual address memory space according to an embodiment. The system includes a plurality of processors 1802a-f that can access the shared memory subsystem 1804. The system incorporates a plurality of programmable hardware memory algorithm processors (MAPs) 1806a-f in a memory subsystem 1804. Each MAP1806a-f may include a memory 1808a-f and one or more field programmable gate array (FPGA) 1810a-f. The MAP provides configurable functional units, and certain algorithms or parts of the algorithms can be provided to the FPGA 1810a-f for processing in close coordination with their respective processors. For example, the MAP can be used to evaluate algebraic expressions for a data model and to perform adaptive data restructuring in embodiments. In this example, each MAP is globally accessible by all processors for this purpose. In one configuration, each MAP can use direct memory access (DMA) to access the associated memory 1808a-f, thereby being independent and asynchronous with each microprocessor 1802a-f. It becomes possible to carry out the task. In this configuration, the MAP can feed the results directly to other MAPs for pipeline and parallel execution of the algorithm.

These computer architectures and systems are just examples, including common processors, coprocessors, FPGAs, and other programmable logic devices, system-on-chip (SOC), application-specific integrated circuits (ASICs), and others. A wide variety of other computer, mobile phone, and personal mobile information terminal architectures and systems can be used in the context of embodiments, including systems that use any combination of processing and logic elements. .. In some embodiments, all or part of the computer system is feasible with software or hardware. Any variety of data storage media includes random access memory, hard drives, flash memory, tape drives, disk arrays, network-attached storage (NAS), and other local or distributed data storage devices and systems. It can be used in connection with the example.

In embodiments, the computer subsystems of the present disclosure are feasible with software modules running in any of the above or other computer architectures and systems. In other embodiments, the functions of the system are firmware, programmable logic devices such as field programmable gate arrays (FPGAs), system-on-chips (SOLs), application-specific integrated circuits (ASICs), or other processing elements and logic. It is partially or fully feasible in the element. For example, set processors and optimizers can be achieved with hardware acceleration using hardware accelerator cards such as accelerator cards.

System Processors and Networks Generally, as shown in the figure, the computer or processor included in the equipment system of the present disclosure is from a medium 11 or network port 05 that can optionally be connected to a server 09 having a fixed medium 12. It can be further understood as a logical device that can read instructions. The system 00 as shown in the figure may include an optional input device such as a CPU 01, a disk drive 03, a keyboard 15 or a mouse 16, and an optional monitor 07. Data communication can be achieved via a designated communication medium for a server at a local or remote location. The communication medium may include any means of transmitting and receiving data. For example, the communication medium can be a network connection, a wireless connection, or an internet connection. Such a connection can provide communication via the World Wide Web. As illustrated, the data relating to this disclosure may be transmitted for reception or viewing by a party 22 over such a network or connection.

The figure shows an exemplary embodiment of a first architectural example of computer system 00 that can be used in connection with the embodiments of the present disclosure. As shown in the figure, an example computer system may include a processor 02 for processing instructions. Non-limiting examples of processors include Intel Xeon ™ processor, AMD Opteron ™ processor, Samsung 32-bit RISC ARM 1176JZ (F) -S v1.0 ™ processor, ARM Cortex-A8 Samsung S5PC100 ( Included is a ™ processor, an ARM Cortex-A8 Apple A4 ™ processor, a Marvel PXA 930 ™ processor, or a functionally equivalent processor. Multithreading of execution can be used for parallel processing. In some embodiments, multiple processors, whether a single computer system with cluster connections or a distributed system with network connections that includes multiple computers, mobile phones, or personal digital assistant devices. Alternatively, a processor with multiple cores can be used.

As shown in the figure, the fast cache 04 can be connected to or mounted on the processor 02 to provide fast memory for instructions or data recently used or frequently used by the processor 02. The processor 02 can be connected to the north bridge 06 by the processor bus 08. The north bridge 06 is connected to a random access memory (RAM) by the memory bus 12 and manages access to the RAM 10 by the processor 02. Northbridge 06 is also connected to Southbridge 14 by the chipset bus 16. The south bridge 14 is then connected to the peripheral bus 18. The peripheral device bus may be, for example, a PCI, PCI-X, PCI Express, or other peripheral device bus. Northbridges and southbridges, often referred to as processor chipsets, manage data transfer between the processor, RAM, and peripheral elements on the peripheral bus 18. In some alternative architectures, the functionality of the northbridge can be installed in the processor instead of using a separate northbridge chip.

In some embodiments, the system 00 may include an accelerator card 22 coupled to a peripheral bus 18. Accelerators may include field programmable gate arrays (FPGAs) or other hardware to accelerate certain processes. For example, accelerators can be used for adaptive data restructuring or to evaluate algebraic expressions used in extended set processing.

The software and data can be stored in external storage 24 and loaded into RAM 10 or cache 04 for use by the processor. System 00 includes an operating system for management system resources; examples of operating systems are, but are not limited to, Linux®, Windows®, MACOS®, BlackBerry OS®. , IOS ™, and other functionally equivalent operating systems, as well as application software running on the operating system for managing data storage and optimization according to embodiments of the invention.

In this example, system 00 also includes network interface cards (NICs) 20 and 21, which are to external storage devices such as network-attached storage (NAS) and other computer systems that can be used for distributed parallel processing. It is connected to a peripheral bus that provides a network interface.

The figure shows an exemplary diagram of network 00 including a plurality of computer systems 02a and 02b, a plurality of mobile phones and personal digital assistants 02c, and a network attached storage storage device (NAS) 04a, 04b. In embodiments, systems 12a, 12b, and 12c can manage data storage and optimize data access to data stored in network-attached storage devices (NAS) 14a and 14b. Mathematical models can be used for the data and evaluated using distributed parallel processing through computer systems 12a and 12b, as well as the entire mobile phone and personal digital assistant system 12c. Computer systems 12a and 12b, as well as mobile phones and personal mobile information terminal systems 12c, can also provide parallel processing for adaptive data restructuring of data stored in network-attached storage (NAS). .. The figure is only an example, and a wide variety of other computer architectures and systems can be used in the context of various embodiments of the invention. For example, a blade server can be used to provide parallelism. Processor blades can be connected via a backplane to provide parallelism. The storage device may also be connected to the backplane or may exist as a network-attached storage device (NAS) via an individual network interface.

In some embodiments, the processor can maintain individual memory space and transmit data to a network interface, backplane, or through other connectors for parallel processing by other processors. In other embodiments, some or all of the processors may use the shared virtual address memory space.

The figure shows an exemplary block diagram of a multiprocessor computer system 00 using a shared virtual address memory space according to an embodiment. The system includes a plurality of processors 02a-f that can access the shared memory subsystem 04. The system incorporates a plurality of programmable hardware memory algorithm processors (MAPs) 06a-f in the memory subsystem 04. Each MAP 06a-f may include memories 08a-f and one or more field programmable gate arrays (FPGAs) 10a-f. The MAP provides configurable functional units and can provide a particular algorithm or part of an algorithm to the FPGAs 10a-f for processing in close coordination with their respective processors. For example, MAPs can be used to evaluate algebraic expressions for data models and to perform adaptive data restructuring in examples of embodiments. In this example, each MAP is globally accessible by all processors for this purpose. In one configuration, each MAP can access the associated memory 08a-f using direct memory access (DMA), thereby independently and asynchronously from the respective microprocessors 02a-f. It will be possible to carry out the task. In this configuration, the MAP can feed the results directly to other MAPs for pipeline processing and parallel execution of algorithms.

These computer architectures and systems are just examples, including common processors, coprocessors, FPGAs, and other programmable logic devices, system-on-chip (SOC), application-specific integrated circuits (ASICs), and others. A wide variety of other computers, mobile phones, and personal mobile information terminal architectures and systems can be used in the context of embodiments, including systems that use any combination of processing and logic elements. .. In some embodiments, all or part of the computer system is feasible in software or hardware. Embodiments of any variety of data storage media, including random access memory, hard drives, flash memory, tape drives, disk arrays, network attached storage (NAS), and other local or distributed data storage devices and systems. Can be used in connection with.

In embodiments, the computer subsystems of the present disclosure are feasible with software modules running in any of the above or other computer architectures and systems. In other embodiments, the functions of the system are firmware, programmable logic devices such as field programmable gate arrays (FPGAs), system-on-chips (SOLs), application-specific integrated circuits (ASICs), or other processing elements and logic. It is partially or fully feasible in the element. For example, set processors and optimizers can be achieved with hardware acceleration using hardware accelerator cards such as accelerator cards.

Although some aspects of the embodiments discussed above will be disclosed in more detail in the following examples, these examples are not intended to limit the scope of the present disclosure in any way.

Example 1
Correction of Single Base Substitution Error This example shows correction of PCR or sequencing error including single base substitution. PCR or sequencing errors involving one-base substitutions have a similar molecular label and the number of occurrences of ≤7 ⁱⁿ the presence of 38 unique probability barcodes ⁽ 17 in the presence of 48 unique probability barcodes). , That is, a copy of the target with the sequencing lead was removed by the step of assigning it as having the same molecular label on multiple targets.

The step of attaching the probability barcode uses a non-depleted pool of 38 ⁽ 6561) unique probability barcodes having oligo (dT) as its binding region and poly (A) in the sample prior to the RT step. It may include the step of labeling the mRNA having. The labeling process may be random and each target molecule can hybridize to one probability barcode. For any target, if the number of target molecules is much smaller than the number of probability barcodes, each target molecule will probably hybridize to a different probability barcode. Therefore, if only a small number of target molecules are present, the small number of target molecules will probably hybridize to a probability barcode with a similar molecular label (ML) during hybridization.

The probabilities of sampling at least one pair of probability barcodes with similar molecule labels from 38 non ^- depleted unique probability barcodes were calculated. Two molecular labels may have similar sequences if they differ by one base. This sampling event can be considered as a sampling involving substitutions, as the probability barcode can actually be non-depleted. This probability can be useful in estimating probability barcodes with similar molecular labels that are least likely to be present for a given sample containing multiple targets. The problem can be articulated as the number of probability barcodes required for at least two probability barcodes with similar molecular labels to be selected with a particular probability. This problem can be articulated as the minimum sample size required for the probabilities of two probability barcodes with similar sequences to exceed 0.5, assuming ³⁸ identifiable molecular labels. can. Therefore, this problem can be thought of as a generalization of the classical birthday problem. The classic birthday problem can assume 365 different birthdays and determine the minimum sample size required for the probability of having two people with the same birthday to exceed 0.5.

To obtain this sample size r, we assume r probability barcodes sampled from ³⁸ unique probability barcodes and use the probabilities of their complement events to have at least one pair of similar molecular labels. The probability was calculated. If only one probability barcode is randomly selected from the ³⁸ unique probability barcodes, the molecular label is not similar to the molecular labels of other probability barcodes because there is only one probability barcode. Probability, p ₁ = 1. If the second probability barcode is also randomly selected from the ³⁸ unique probability barcodes, the probability that the molecular label is not similar to the molecular label of the first probability barcode, p ₂ = ⁽ 38-). 16-1) / ³⁸ . This assumes that there are three possible bases at each position in the probability barcode, and for a given molecular label, each base position has two possible alternative nucleotides, resulting in a total of 2 *. This was because eight 1-base variants were obtained. If the third probability barcode is continuously and randomly retrieved from the ³⁸ unique probability barcodes with the unique molecular label, the probability that the molecular label is not similar to the previous two molecular labels, p ₃ = (3 ⁸ -1-16-1-16) / 3 ⁸ = ( ^{3 8} -2 * 17) / 38. The probability barcode can be continuously extracted from the 38 unique probability barcodes up to the ^rth probability barcode. This last probability barcode is a probability that is not similar to the previous probability barcode, pr (38- ( _r - ¹ ) * 17) / ³⁸ . Since all r probability barcodes were taken out independently, the probability of taking out probability barcodes that do not have similar sequences is P (all molecule labels that do not have similar sequences) = p ₁ * p. ₂ * p ₃ * ... _pr . Therefore, the probability of having at least one pair of similar probability barcodes among r probability barcodes from 38 probability barcodes with unique molecular labels is P ⁽ at least one pair of molecules with similar sequences). Label) = 1-P (whole molecule label without similar sequence). The sample size is then set from this equation by setting the desired value = 0.01, 0.05, 0.1, or the desired value for P (at least one pair of molecular labels with similar sequences). r was calculated.

Table ¹ shows the probabilities of having at least one similar pair between r molecular labels, assuming 38 or ⁴⁸ unique molecular labels. If a unique probability barcode of 38 and a probability barcode of ^{≤7 (} 17 if there is a unique probability barcode of 48) are selected, a pair of probability barcodes with similar molecular labels will be used. The probability of observing is less than 0.05, which is negligible. Thus, as justified by this small probability, similar molecular labels were more likely to be man-made than real prospective selections of similar probability barcodes and can be corrected.

However, if there are more than 7-24 probability barcodes, the probability of observing one or more pairs of probability barcodes with similar molecular labels will be high (eg 0.5). Therefore, we cannot be confident that these probability barcodes are true and non-artificial. In contrast, the general intuition is that if only 24 probability barcodes were retrieved from the 6561 unique pool, any one base deviation could be the result of a sequencing error rather than a coincidence. Would have been erroneously concluded.

For example, if 115 probability barcodes are randomly sampled, there is only one probability calculated, so it is 100% certain that there is at least one pair of probability barcodes with similar molecular labels. .. Assuming 115 targets in the sample, after hybridization and reverse transcription processes, two pairs of probability barcodes with similar molecular labels and 111 probability barcodes with dissimilar molecular labels (total). 115 probability barcodes) become observable. However, if three pairs of probability barcodes with similar molecular labels and 110 probability barcodes with dissimilar molecular labels (a total of 116 probability barcodes) are observed in the sequencing data, Only two pairs of probability barcodes with similar molecular labels are true, and the third pair may have been generated by some error. This 100% probability is that the event of observing at least one pair of probability barcodes with similar molecular labels can occur if 115 probability barcodes are randomly sampled during the probability barcode attachment process. However, this does not mean that all observed pairs of similar molecular labels are true. Probability barcodes with similar molecular labels can be generated from probabilistic bar coding steps, real or true molecular labels, or from PCR errors, man-made objects, or sequencing errors, errors or false molecular labels. Therefore, if similar molecular labels are observed, further evaluation may be needed to determine if a particular pair of molecular labels is true. ^In addition, if the ^total molecular label diversity is increased from 38 to 48, additional probability barcodes may be needed for each probability to predict similar pairs of molecular labels.

Tables 2 and 3 show that when a probability bar code of ≤7 with a unique molecular label was observed, the probability of such occurrence was less than 0.05, so it is highly probable that similar molecular labels will occur. Indicates that it was low. Therefore, such similar molecular labels may have been caused by PCR errors, man-made objects, or sequencing errors, which should be removed from the molecular label counts in order to correct or adjust the molecular label counts. .. Thus, the total number of true molecular labels in Tables 2 and 3 can be reduced from 5 to 1 and from 7 to 6, respectively. However, in Table 4, 23 unique barcodes have been observed, which predicts an approximately 50% chance of having at least one pair of probability barcodes with similar molecular labels. Therefore, 16 pairs of probability barcodes with similar molecular labels may be real, and each pair of similar molecular labels will undergo further evaluation to see if they are real or not. It will take.

Overall, from these data, the number of probability barcodes with similar molecular labels observed is likely due to PCR errors, man-made objects, or sequencing errors. Since it occurred, it is clear that it was removed.

Example 2
Determining the Quality Status of Targets in Sequencing Data This example shows that the quality status of a target in sequencing data is a complete sequencing quality status, an incomplete sequencing quality status, or a saturated sequencing quality status. Clarify the decision process. The quality status of the target depended on whether all of the true or real molecular labels were observed.

As shown in Example 1, the complete counting of probability barcodes with unique molecular labels present in the library can be highly dependent on the sequencing depth. The deeper the sequencing, the more likely it is that all true molecular labels will be observed. Shallow sequencing, while inexpensive, can miss a large number of molecular labels and possibly impair gene detection sensitivity. Complete sequencing means that all true molecular labels of the probabilistic bar code used to label the target molecule have been observed, and incomplete sequencing means that only part of the true molecular label has been observed. It can mean that. In addition, more than 48568 target molecules could be present in the starting sample, which is the lower limit of Poisson-corrected or regulated number of molecules based on the recognizable probability bar code 6561-2 * standard deviation. be. Second, saturated sequencing can occur when the number of target molecules is difficult to determine due to limitations on the diversity of all molecular labels. However, when a small amount of RNA is used as an input for probabilistic barcodes, saturation sequencing is less likely to occur.

To mathematically define complete or incomplete sequencing, each was compared to a theoretical model with no errors. Under full experimental conditions, each copy of the target molecule in the starting sample can generate (1 + C) ^j copies, assuming C efficiency in the jPCR cycle and each cycle. For each barcoded molecule in the starting sample, illumination sequencing can be considered as Poisson sampling from (1 + C) ^j clonal copies amplified from the original barcoded molecule. Theoretically, for the same target gene, sequencing of k target molecules with probability barcodes can be evenly expressed after PCR for all molecules with probability barcodes, so (1 + C) ^j copies. Can be considered as an iterative Poisson sampling from. An important assumption of the Poisson model was that the mean should be equal to the variance and the sequencing leads should follow equal dispersal. Spraying can be defined as variance / average.

In fact, full sequencing usually involves clustered errors with a much lower read frequency. Unlike true molecular labeling, errors are unlikely to participate in every PCR cycle, resulting in much greater read frequency changes compared to Poisson. 19A-19B show examples of complete and incomplete sequencing genes. In FIG. 19A, the maximum sequencing lead exceeded 350 times the minimum sequencing lead. Therefore, complete sequencing tends to show a larger dispersal index (> 1) than Poisson.

In contrast, in the case of incomplete sequencing, the change in sequencing reads is small compared to Poisson because only part of the probability barcode with the true molecular label in the library was sequenced. In FIG. 19B, the maximum sequencing lead was only about three times the minimum sequencing lead. Therefore, incomplete sequencing tends to show a smaller spray index (<1) than Poisson.

Besides calculating the dispersal index, the most abundant molecularly labeled sequencing leads can be used to determine if the sequencing is complete. For example, if the read for abundant molecular indicators is 25 and the dispersal index is 5, the sequencing status can be classified as complete; otherwise, it can be classified as incomplete. Since sequencing may be incomplete until sequencing errors begin to appear, a 25 read threshold can be used. Sequencing errors can occur if any molecular label is found more than 25 times.

Highly abundant gene sequencing data is saturated in the probability barcode, for example, in the case of a 38 probability barcode with a unique molecular label, other underexpressed genes ⁱⁿ the same well under conditions above 6557. Sequencing information can be used to calculate the dispersal index and maximum sequencing read for that gene. For example, if the second most abundant gene in the same well is not saturated in the probability barcode and is classified as incomplete sequencing, then the saturation of the first gene can be considered real. , The number of molecules cannot be calculated. And if the second abundant gene is classified as complete sequencing, the saturation of the first gene may be artificial and the appearance of all probability barcodes may be due to an error. .. A Poisson model-based thresholding algorithm can then be used to determine the number of true molecular labels.

Overall, these data reveal the process of determining whether the sequencing status is complete sequencing, incomplete sequencing, or saturated sequencing.

Example 3
Correction of PCR or Sequencing Errors by Single Base Substitution of Fully Sequencing Genes This example is a single base for a fully sequenced gene, i.e., a gene with a complete sequencing quality status in the sequencing data. The steps of correcting PCR or sequencing errors due to substitution are shown. This example also shows the step of thresholding the molecular label of a target, eg, a gene, in order to determine the true and false molecular labels associated with the target in sequence data.

Sequencing error rates per nucleotide can vary from 0.1 to 1% and can usually be seen as infrequent reads. As the sequencing progresses deeper, more sequencing errors can occur. For example, if the true nucleotide sequencing error is 0.5% and the molecular label is sequenced 100 times, the predicted number of sequencing errors associated with this molecular label is that the molecular label is 8 nucleotides long. If there is, it can be about 4 calculated from 100 * (1- (1-0.5%) ⁸ ). If the molecular label is sequenced 300 times, the predicted number of sequencing errors can be about 12. These sequencing errors can generate artificial molecularly labeled sequences that can increase the count. These molecular labels can be removed prior to further analysis.

Of all sequencing errors, single-base errors can occur much more often than those separated by two or more bases. The probability of having a single nucleotide sequencing error can be derived from the binomial distribution including the sample size 8 and the probability of success equal to the single nucleotide sequencing error rate. One goal was to correct single base sequencing errors. Single-base sequencing errors can be thought of as children of the most abundant and close (eg, Hamming distance) molecular labels, ie parent molecule labels. Sequencing errors were detected by finding the true child of the parent molecule label (ie, the child molecule label one base away from the parent molecule label).

Selection of Parent and Child Molecular Labels Parent molecule labels may be required to have> 25 sequencing leads and child molecule labels may be required to have 3 or less sequencing leads. These requirements were based on the following reasoning. It is assumed that the probability of sequencing error for each nucleotide is 0.5%. If the molecular label was sequenced 25 times to produce a total of 200 nucleotides, then 200 * 0.005 = 1, so 1 nucleotide was expected to be an error. Therefore, for each molecular label with 25 sequencing leads, it was expected to have at least one child molecular label. It can be assumed that the parent molecule label should have 25 sequencing leads. A child molecule label with a sequencing lead of 4 was less likely to be a sequencing error. This was because the probability of introducing the same error four times in one molecular label was 8 * 0.005 ⁴ = 10 ^-9 . If a total of 106 sequencing error reads were present, the expected number of sequence errors repeated four times was 5 * 10 ⁹ * 10 ⁶ = 0.005, which could be ignored. Therefore, the child molecule label should have a lead ≤ 3.

Assuming a parent molecule label separated by one base and its associated child molecule label, how to determine the child molecule label that is the true sequencing error of the parent?
Using a multiple binomial test, assuming a set of child molecule labels that have a parent molecule label and a sequencing read (R _child1 , R _child2 , ..., R _childm ) that are one base different from the parent molecule label. The true child molecule label can be identified. Under the null hypothesis, the abundance of true child molecule labels should be less than or equal to R _par * p (mathematical, H ₀ : p <e / 2); otherwise, the abundance is , R _par * p is greater than (HA: p <e / 2), and the hypothesis that the molecular label was a true child molecular label can be rejected. Next, the probability that a child molecule label different from the parent molecule label by one base is observed once is p = e / 2. Next, mathematically, from the total abundance (R _child + R _par ), the probability of observing this _child molecule label at least R _child times is as follows:

If the child molecule label was actually a sequencing error of its parent molecule label, the probability p _child should be greater than the critical value of 5%. Since multiple hypotheses are tested simultaneously, the critical value used to reject the null hypothesis can be determined by false discovery rate (FDR), which is controlled at the 5% level, and p _child is at the 5% level. If it is larger than the FDR of, the hypothesis can be accepted. With a 5% controlled FDR, unadjusted p-values can be sorted in ascending order, for example p ₁ ≤ p ₂ ≤ p _m . Next, a test including the corresponding rank j can be found. If p _child ≤ j / m * 5%, the null hypothesis that this child molecule label was a one-base sequencing error of the parent molecule label can be tolerated.

Overall, these data demonstrate a step of correcting single-base sequencing errors for fully sequenced genes: step (1), the richest sequencing read if the sequencing read is greater than 25. The molecular label having the above is selected as the first parent molecule label. Step (2) Select molecular labels with sequencing leads ≤ 3, identify these molecular labels that differ by 1 base from the first parent molecule label, and call them child molecule labels; child molecule labels or 1 base. If no child molecule label is found, the process proceeds to step (5). Step (3), multiple binomial tests are performed on all of the child molecule labels and parent molecule labels, the child molecule labels for which the null hypothesis is acceptable are removed, and their sequencing leads are labeled with their parent molecule labels. Attribut to. If none of the null hypotheses were accepted, this means that not all child molecule labels were single-base sequencing errors of the parent molecule label, and no read corrections need to be performed. Step (4), molecularly labeled sequences and sequencing reads are updated. Step (5): Select a molecular label with the next largest sequencing lead as the parent molecule label and repeat the above steps until no eligible parent molecule label or qualified child molecule label remains.

Table 5 shows updated TFRC sequencing data after removing single base sequencing errors using the analysis described above. The unique number of molecular labels decreased from 23 (shown in Table 4) to 11.

Using Poisson Models for Thresholding Sequencing errors can be more likely to occur under full sequencing. Some types of errors, such as single nucleotide sequencing errors, can be corrected, while other errors, such as random integration of artificial molecular labels, will not be correctable based on sequence similarity. Instead, these types of errors can be identified by modeling. As mentioned earlier, complete sequencing tends to be oversprayed for Poisson. Therefore, two unique Poisson models characterized by overspraying were created: one is true molecular labeling (ie, the molecular labeling sequence used to label the target molecule during the probabilistic bar coding process). A second model can be used to model a sequencing read for an error molecule label (ie, not used during the probability bar coding process, but appears after sequencing due to an error). Can be used for molecularly labeled sequences). The sequencing error rate can be about 0.1-1% and the PCR cycle error rate can be about 0.001%. PCR errors can occur more often during the post-PCR cycle, resulting in error molecular labeling with low sequencing reads, but can contribute to the majority of all observed molecularly labeled sequences. Therefore, errors caused by PCR and sequencing can often have lower sequencing reads than true molecular labeling. Therefore, the Poisson average of the sequencing read of the true molecular label is larger than the Poisson average of the error molecular label.

（式中、Ｐ（Ｘ_i＝Ｒ_i｜μ_t）は、平均μ_tを有するポアソン過程の下で、存在量Ｒ_iを有するｉ番目の分子標識を観察する確率を示す）。 There are a total of k identifiable molecular labels, t of which are true molecular labels such as BC ₁ , BC ₂ , ..., BC _t , and the rest are BC _{t + 1} , BC. Suppose it was an error molecule label such as _{t + 2} , ..., BC _k . Sequencing leads mapped to these true and error molecular labels can be R ₁ , R ₂ , ..., R _t and R _{t + 1} , R _{t + 2} , ..., R _k . .. Furthermore, assuming that the Poisson mean with true and error molecular labels is μ _t and μ _n (μ _t > μ _n ), the overall process probabilities are as follows:

(In the equation, P (X _i = R _i | μ _t ) indicates the probability of observing the i-th molecular label with abundance R _i under a Poisson process with mean μ _t ).

To determine the number of true molecular labels, we considered the number of models as follows; starting from the model assuming that all molecular labels are true (hence, l = k); the least. The second model assuming that the molecular label is an error and all other molecular labels are true (hence l = k-1); only the most abundant molecular labels are true and everything else is true. Up to the last model assumed to be an error molecule label (hence l = 1). Finally, the best model is equivalent to the Akaike Information Criterion (AIC), which has the highest or lowest likelihood among all the models considered, and the AIC is for a given data. It can be used for model selection by measuring the relative quantity of each of the possible models. Mathematically, AIC is defined as AIC = -logL + 2p, where p is the number of parameters estimated by the model. Therefore, for L _k and L ₁ , p = 1, otherwise p = 2. From the examples shown in Table 6, it can be seen that of the eight possible models compared, only the three molecular labels with the three largest sequencing leads are considered to be true molecular labels. Also, FIG. 20 shows that the thresholds derived from the selected model (the three largest) clearly distinguished true molecular labels from those with a high probability of error.

The data show sequencing reads of fully sequenced genes corrected by removing single-base sequencing errors and thresholding using a Poisson model.

Example 4
Regulation of Incompletely Sequencing Genes This example removes the noisy gene and uses a zero-cleavage Poisson model to estimate the total number of molecular labels expected to be present in the library. The process of regulating an incompletely sequenced gene is shown.

Removal of Noisy Genes Besides considering the statistics of molecular labeling and its sequencing reads, gene-level analysis can also be useful. A gene can be considered noisy if there are very few molecular labels detected for a gene and each molecular label has a significantly lower read than a fully sequenced gene. This assumption was based on the argument that molecules with probability barcodes in the same library should be amplified and sequenced at approximately the same frequency. These expectations can be influenced by PCR and sequencing bias due to differences in sequencing of each molecule, which are caused by events such as sample contamination and unwanted molecular recombination during PCR. It was supposed to be small accordingly. A gene can be noisy if its amplification rate (mean read per molecular label) is similar to the amplification rate of errors derived from fully sequenced genes in the same library.

であった。同様にして、他の完全にシーケンシングされた遺伝子すべてのｇ₂、ｇ₃、・・・、ｇ_xについて、ＥＡＭＰを計算することができる。観察された計５未満の分子標識を有する潜在的ノイジー遺伝子ｇ’₁、ならびに各分子標識にマッピングされたＲ_g’1,1、Ｒ_g’1,2、・・・、Ｒ_g’1,kシーケンシングリードについて、カットオフを適用することができ、その増幅速度を

として決定する。ａｍｐ_g’1＜中央（ａｍｐ_g’1、ａｍｐ_g’2、・・・、ａｍｐ_g’x）であれば、遺伝子ｇ’₁をノイジー遺伝子であると考えた。そうでなければ、これは、不完全遺伝子とみなすことができる。同様に、他のノイジー遺伝子も検定し、除去した。５の分子標識をカットオフとして選択した理由は、低い増幅速度を有する遺伝子を２つの個別のケース：人工物（５未満の分子標識が観察されたもの）と不完全シーケンシング（低ＰＣＲ／シーケンシングのプライマー失敗により≧５の分子標識が観察されたもの）に処理することが望ましいと思われるためである。 Specifically, the fully sequenced gene g1 is composed of a total of t ₁ true molecular labels and e ₁ error molecular labels, thereby R _g1,1 and R _g1,2 , ..., R _{g1, t1} are the sequencing leads mapped to the true molecular label, and R ^* _g1,1 , R ^* _g1,2 , ..., R ^* _{g1, e1} are the error molecular labels. Suppose it is a sequencing read mapped to. Next, the amplification rate (EAMP) of the error molecule label of g ₁ is

Met. Similarly, EAMP can be calculated for g ₂ , g ₃ , ..., G _x for all other fully sequenced genes. Potential noisy genes g'1 with a total of less than 5 molecular labels observed, as well as R g'1, ₁ , R _g'1 , 2, ..., R _{g'1 ,} mapped to each molecular label. A cutoff can be applied to the _k -sequencing read to determine its amplification speed.

To be determined as. If amp _g'1 <center (amp _g'1 , amp _g'2 , ..., amp _g'x ), the gene g'1 was considered to be _a noisy gene. Otherwise, this can be considered an incomplete gene. Similarly, other noisy genes were tested and removed. The reason for choosing the molecular label of 5 as the cutoff is that genes with low amplification rates were selected in two separate cases: artificial (where less than 5 molecular labels were observed) and incomplete sequencing (low PCR / sequencing). This is because it seems desirable to treat the molecule labeled with ≧ 5 due to the failure of the primer of the sing.

When estimated sequencing using the zero-cut Poisson model was incomplete, errors could still be present in the data, but could be difficult to identify due to inadequate sequencing reads overall. If the sequencing is shallow and not all of the molecular labels present in the library are observed, some assumptions may be needed for important analysis. It can be assumed that all observed molecular labels are true, and that true unobserved molecular labels are cleaved at zero, i.e., cleaved molecular labels observed at zero time. .. Not all transcripts with probability barcodes are sequenced for a given gene, but all live by applying a zero-cleaving Poisson model using the frequency of detected molecularly labeled reads. The complete variety of molecular labels present in the rally can be estimated.

K identifiable molecular labels with leads (R ₁ , R ₂ , ..., R _k ) were observed, no (SK) molecular labels were observed, and the leads were zero. One goal was to estimate S, the total number of molecular labels expected to be present in the library. Assuming a frequency of recognizing sequencing reads 1, 2, 3, or more as Poisson variates cleaved at zero by the Poisson mean μ, if the sum of all sequencing reads is n, the likelihood is It can be expressed as:
L (S, μ) ∝S! / (S-k)! μ ⁿ exp (-Sμ) (Equation 3)

Traditional reasoning methods can be applied to estimate μ, S and their standard errors. The maximum likelihood (MLE) of μ is n / S, and the approximation of S to MLE is k / (1-e- ^{n / S} ) or k / (1- (1-1 / S) ⁿ . ) Can be. FIG. 21 shows a fitted zero-cut Poisson model based on the number of molecular labels and their corresponding sequencing leads. As shown in FIG. 21, 33 unique molecular labels were observed across a total of 39 leads in the partially sequenced library. Based on the frequency of molecular labels with sequencing reads 1, 2, 3, and 4, a Poisson model was applied to estimate that a total of 113 molecular labels in the entire library had advanced sequencing to completion. did. The estimation method was applied for estimation of μ, S and their standard error. The MLE of μ is n / S, and the approximate value of S to MLE can be k / (1-e- ^{n / S} ) or k / (1- (1-1 / S) ⁿ ).

Overall, these data are incomplete sequences corrected by removing the noisy gene and using a zero-cleaving Poisson model to estimate the total number of molecular labels expected to be present in the library. Clarify the sequencing reads of the single gene.

Example 5
Complete Sequencing Genes and Incomplete Sequencing Genes This example shows an example of the output produced after regulating the sequencing reads of the complete sequencing genes and incomplete sequencing genes.

Table 7 provides an example of the output produced after regulating the sequencing reads of complete and incomplete sequencing genes. The description of the column headings was as follows: "Gene ID" indicates the name of the detected gene. The "sequencing status" indicates three possible outcomes: complete, incomplete and saturated, which determine the method of analysis. Classification was performed according to the spray index and the sequencing leads mapped to the most abundant molecular labels (ML). "Uncorrected ML" indicates the count of unique molecular labels observed for that gene ("0" for undetected genes). “Uncorrected read” indicates the total number of sequencing reads mapped to the uncorrected ML (“0” for undetected genes). The corrected ML shows the count of unique molecular labels that were considered true molecular labels after applying the algorithm (only for fully sequenced genes, "NA" for incomplete genes, noisy and undetected genes). Is "0"). "Corrected read" indicates the total number of sequencing reads mapped to the corrected ML ("NA" for complete sequencing genes only, "NA" for incomplete genes, "0" for noisy and undetected genes). .. "Extrapolated ML" indicates the estimated number of unique molecular labels by the zero-cleaving Poisson model ("NA" for incompletely sequenced genes only, "NA" for complete genes, "0" for noisy and undetected genes. ). "Estimated Mol" indicates the number of molecules estimated based on the corrected ML (for complete sequencing genes) or supplemental ML (for incomplete sequencing genes), and for noisy and undetected genes. It is "0". "Estimated Mol LB" indicates the lower limit of the estimated number of molecules. "Estimated Mol UB" indicates the upper limit of the estimated number of molecules.

を用いて計算した。 In Table 7, the estimated Mol (n), which is the estimated number of starting molecules, was calculated as follows:
n = -mlog (1-k / m), formula (4)
(In the formula, m was the total diversity of molecular labels ( ³⁸ ) and k was the total number of unique molecular labels observed). The variance of n, var (n), was derived using the Taylor expansion: var (n) = (m / (m—k)) ² var (k) (in the equation, var (k) is m. * Can be expressed as (1- (1-1 / m) ⁿ ) (1-1 / m) ⁿ + m (m-1) ((1-2 / m) ^n- (1-1 / m) ²ⁿ ) can). The lower and upper limits (estimated Mol LB and estimated Mol UB) of the estimated number of starting molecules are

Was calculated using.

Overall, these data reveal steps to regulate complete and incomplete sequencing genes.

Example 6
Performance of Correction of Completely Sequencing Genes and Incompletely Sequencing Genes This example demonstrates the performance of correcting sequential reads of fully sequenced genes. This performance was based on the error and noise of the removed uncorrected molecular label count and the removed sequencing read.

Several fully-sequencing genes were selected to test the ability of the complete-sequencing gene to correct the sequencing reads. Table 8 compares several measurements for these genes before and after correcting or regulating sequencing reads. Uncorrected ML, uncorrected read, corrected ML, and corrected lead were introduced directly from the output table. Uncorrected amp (amplification rate using uncorrected data) and filtering amp (amplification rate using corrected true molecular labeling data), (uncorrected read / uncorrected ML) and (corrected read / corrected ML) Was calculated using. After correction of the total number of observed molecular labels, the percentage of retained ML to the number of true molecular labels is 100 * corrected ML / uncorrected ML, and the retained% read is also 100 * corrected. Defined as lead / uncorrected lead. Table 8 shows examples of genes at various abundance levels, including GAPDH and ACTB that exhibit more molecular labels and total reads. The number of true molecular labels after applying the corrections accounted for less than 7% of the total molecular labels found in the uncorrected data, which is considered to be more than 93% of the molecular labels as error molecular labels. It means that it was discarded. 93% of the uncorrected molecular labels were removed as noise, but true molecular labels contributed to at least 72% of the leads, which means that these discarded error molecular labels consist of much lower leads. It also means that. Moreover, the amplification rate after applying the algorithm was in the range of 137-413, which was much higher than that obtained using the uncorrected data (6.1-29.4). The corrected amplification rate was a much more practical measurement, which correlated with a PCR efficiency of at least 75%.

Overall, these data significantly reduce errors and noise in uncorrected molecularly labeled counting data, while correcting the sequencing reads of the complete sequencing gene, while still maintaining the ability to use most of the sequencing reads. Indicates that it has decreased.

Example 7
Tools for Summarizing and Visualizing Probabilistic Barcoded Target Counting Data This example presents a tool for summarizing and visualizing probabilistic barcoded target counting data as shown in the previous embodiment.

Two plates of single cells were made for treatment with Precision ™ assay (Cellular Research, Inc. (Palo Alto, CA)) for assay data. Two different cell types were used in this experiment in a 4: 1 ratio, and the identity of the cells placed in each well was obscured by the researchers conducting the experiment. The goal of this study was to identify the cell type of each well using the gene expression profile from the probability barcode count.

In order to evaluate the overall sequencing data quality in the wells, the sum of the sequencing reads for each well was calculated. Then, in order to evaluate the performance of the correction method, some statistical measurements before and after the application of the correction method were aggregated and compared. In addition, graphs provide a visual representation of the data and can easily detect anomalies or patterns.

9 and 10 show the total number of sequencing leads per well of plate 1 containing sequencing leads <5000 (italics). Wells with much lower leads, such as leads <5000, may indicate that no single cell was assigned to the wells and should be excluded for further analysis.

Tables 10 and 11 compare several measurements before and after the correction method. From these tables, large fluctuations were observed in "uncorrected leads" (total of sequencing reads per well) and "uncorrected ML" (total number of molecular labeling counts per well). This large variation may be due to their standard deviation (SD) being greater than average, which is also an indication of the presence of low lead wells. After using this method, about 47% of the genes per well were classified as complete sequencing genes among all the genes present. If the majority of the genes are classified as incomplete sequencing genes (eg 0%), this method may not remove noise in the data. For each well, approximately 15% molecular labeling was retained after complete gene correction, but these molecular labels were mapped to an average of 95% sequencing leads. The higher the value of the% read retained, the more effectively the correction method can capture the signal (lead given from the true molecular label) while removing noise. Also, the amplification rate of each molecular label retained as a true molecular label was 163.32, much higher than 22.76 before applying the correction method.

FIG. 22 shows a bar graph of total sequencing leads per well. FIG. 22 achieves direct visualization of relative inputs across 96 wells. From this figure, it can be seen that wells C02 and F11 have higher leads than others, which may indicate multicellularity for these wells. Wells A12, B01, B07 to B12, C03, C04, C07, C11, D07, D08, D11, E05, E08, F04 to F10, F12, G03, G07, H03, H04, H07 to H09, H10 to H11 It has a much lower lead compared to the other wells, which may indicate that no cells were placed in these wells.

FIG. 23 shows a bar graph of the% complete sequencing gene, the% molecular label (ML) retained as a true molecular label, and the% retained lead mapped to the ML retained for each well. FIG. 23 shows the percentage of genes classified as complete per well for which correction methods can be applied to remove noise (bottom of each well); the level of noise per well with molecular labeling (the bottom of each well). The percentage of molecular labels that were considered true molecular labels after the application of the correction method, as opposed to the molecular labels that were observed prior to the application of the correction method); and per well with sequencing leads. No noise level (percentage of leads mapped to true molecular label compared to all uncorrected leads, middle row of each well). As shown, the percentage of complete sequencing genes varies with wells, but wells A12, B01, B07-B12, C03, C04, C07, D07, D08, D11, E05, E08, F04-F10, F12. , G03, G07, H03, H06, H07, H10-H11 were much lower, consistent with wells with much lower leads. The% retention ML indicated by the upper row was generally less than 20% in all wells, while the% retention lead indicated by the middle row exceeded 90% in all wells. This type of plot can provide a concept of how effective the correction method is in removing noise and, on the other hand, maximizing the signal in each well.

FIG. 24 shows a boxplot of% retained leads that varies genetically for each well. The boxplot at the gene level reveals detailed information such as how well the correction method worked for each gene in the well, which cannot be represented by a bar graph at the well level. From the boxplot of% retained reads for all fully sequenced genes per well shown in FIG. 24, the variation between genes is, for example, wells D11, F4, F8, H3 and H8 with whiskers greater than 0.6. It turns out that it can be important, such as in the case of. However, these five wells corresponded to much lower total sequencing leads, 3357, 5457, 2874, 3414 and 4043.

Clustering can be used to analyze gene expression data. Principal component analysis (PCA) can be used for dimensionality reduction by reducing multidimensionality and possibly correlating variables to a small number of variables by orthogonal transformation. The main principal components from the PCA can be used to retrieve the clusters in the data.

25A-25B show PCA plots with uncorrected ML vs. algorithm applied corrected MI from two plates. FIG. 25A shows a PCA plot with uncorrected ML per gene per well with total sequencing reads> 5000. This PCA plot first removes wells with total sequencing leads <5000 (resulting in 139 wells and 107 genes remaining, excluding 3 regulatory genes); second. Steps of removing genes with zero uncorrected ML across 139 wells (85 genes remained); third, adopting the uncorrected ML plus one log and incorporating zeros into the dataset, Generated by the process of applying PCA to log data after centering and scaling. The PCA plot clearly shows two clusters, but for wells such as D02, D05, and F06, which are approximately equal distances from both clusters, cell typing was difficult. The result of clustering can be compromised due to the addition of noise, and even a small number of noise variables can compromise a well-defined cluster structure. Therefore, it can benefit from feature / variable selection pretreatment steps or filtering or denoising steps. By applying the correction method to the complete sequencing data, a clear cluster structure was achieved, as shown in FIG. 25B. The PCA plot in FIG. 25B is not the uncorrected ML (count of molecular labels of all genes detected prior to application of the algorithm), but the corrected ML (count of true molecular labels of fully sequenced genes after applying the correction method). ) Was used, a clear cluster structure was obtained as shown in FIG. 25A, and 75 genes were used in a total of 139 wells. Two distinguishable clusters were observed, which were successfully separated by the y-axis (PC2). Compared to FIG. 25A, the cluster of FIG. 25B was compact in size and cells in each well were clearly assigned to the cluster. In addition, the small cluster on the right side of the y-axis in FIG. 25B consists of 31 wells, which is about 22% of the total wells, much closer to the expected 20%.

Overall, these data reveal some useful tools for summarizing and visualizing data counting of targets with probability barcodes.

Example 8
ML coverage of each ML in a plate of highly expressed gene-ACTB In this example, the discriminating distribution of ML errors that occurred during sequencing or PCR generally has a discriminating distribution from ML. To demonstrate.

In addition to absolute gene expression counting and PCR bias correction, ML can provide a better understanding of library fabrication methods and statistical quality of sequencing data. With respect to the number of reads showing the same gene ML (called ML coverage), it is possible to detect sequencing error base calls or PCR errors generated during library production. For example, the gene ML from a given SL represented by multiple reads is probably an accurate measurement as compared to the gene ML from a given SL represented by a unit read alone. Low ML coverage barcodes in the presence of high ML coverage in the same library are often man-made objects or errors generated during sequencing runs or PCR steps during library fabrication. ML errors that occur during sequencing or PCR generally have a discernible distribution from true ML. FIG. 27 shows an exemplary plot showing the molecular label coverage of each molecular label in the highly expressed gene-ATCB microplate, where the discernible distribution is observed between the error molecular label and the real molecular label. Was done. FIG. 28 is an exemplary plot showing the fit of the two negative binomial distributions to the molecular label coverage of each molecular label in a highly expressed gene-ATCB microplate. The fit of the two negative binomial distributions demonstrates that the molecular labeling error with a low molecular labeling depth and the true molecular labeling with a higher molecular labeling depth are statistically distinguishable distributions. The x-axis is the molecular depth.

Overall, these data demonstrate that ML errors that occur during sequencing or PCR generally have a discernible distribution from true ML.

Example 9
Correction of Molecular Labeling Due to PCR or Sequencing Errors This example reveals a method for correcting molecular labeling due to PCR and sequencing substitution errors, which is complete and without assumption of uniform coverage. It can be applied to all transcriptome assays without the need for high sequencing coverage due to sequencing status.

Deduplication was performed on the first mapping coordinates and unique molecular label (UMI) of each lead, assuming the same starting coordinates, UML, and strands, assuming the leads were identical. After deduplication, the UML with the highest count per cluster was retained (Table 13).

The molecular label (ML) was corrected for each gene. For each gene, ML clusters were identified using directional proximity. If the ML was within a Hamming distance of 1 and the parent ML count ≥ 2 * (child MI count) -1, the directional proximity method clustered the ML. All MLs in the same cluster were considered to be derived from the same parent ML, and the child ML count was collapsed to the parent ML. FIG. 29 shows the molecular labeling correction, where the pairwise Hamming distance of 1 occupies a large proportion. After correction of the molecular label, molecular labels with different Hamming distances of 1 were clustered and folded into the same parent molecule label. FIG. 30 shows a curve of the corrected number of MLs with respect to the lead number coverage. Since all reads were retained, this method can also be used to eliminate single-base PCR or sequencing errors.

Overall, these data demonstrate a correction method that can be applied to correct or adjust the data for the entire transcriptome assay since all reads were retained.

Example 10
Molecular Labeling Counting for High Input Samples This example illustrates the unique molecular labeling used when the number of input molecules increases.

The BD Precise ™ Targeted Assay is considered most suitable when used for small sample inputs (eg, single cells) to allow for mRNA probabilities and unique labeling. As the number of transcripts increased relative to the barcode pool in high RNA / cell input experiments, the percentage of ML that was minimized to label the same gene increased, which was theoretically calculated using the Poisson distribution (Figure). 26). Under these circumstances, the step of quantifying gene expression using ML without statistical correction underestimates the number of initially present molecules without Poisson correction or correction based on two negative binomial distributions. Will do.

Poisson corrections or corrections based on two negative binomial distributions are no longer possible in extremely high input samples where the number of mRNAs per gene exceeds the entire collection of 6651 barcodes. For example, a maximum of 6651 saturated barcodes are expected in either case, regardless of whether it is a 65,000 or 100,000 input molecule. Thus, genes and samples that appear to have high sample inputs can be modified, whereby the ML count will probably be underestimated.

Overall, these data demonstrate the need to adjust uncorrected data when quantifying gene expression using ML.

Example 11
Recursive Replacement Error Correction (RSEC)
This embodiment reveals recursive replacement error correction.

Two collaborative methods can be used in the BD Precision ™ Targeted Assay analysis pipeline to eliminate ML errors. Briefly, ML errors resulting from sequencing base call substitution errors are identified and adjusted to true ML barcodes using recursive substitution error correction (RSEC). Subsequently, ML errors or sequencing base deletion errors from the library fabrication process are adjusted using distribution-based error correction (DBEC).

The RSEC algorithm can regulate ML errors resulting from PCR or sequencing substitutions. These rare error events are acknowledged when examining ML coverage. For example, the ML coverage of error ML can be significantly lower than that of ML in a suitable sequencing sample (FIG. 27); When used, they generally have similar ML coverage and do not need to be removed. RSEC can be important for adjusting the ML count of libraries with advanced sequencing barcodes, as more ML errors will appear as the sequencing depth increases.

Briefly, RSEC considers two factors in error correction: 1) ML sequence similarity; and 2) their ML coverage. For each target gene, the MLs are connected if both of their ML sequences are within one base (Hamming distance = 1) with respect to each other. For each connection between MLx and y
Coverage rate (y)> 2 * Coverage rate (x) + 1 equation (5)
(In the formula, y indicates "parent ML" and x indicates "child ML").

Based on this substitution, the child ML can be folded into its parent. This process is recursive until there are no more identifiable parent / child MLs for the gene.

は、

へ折りたたまれうる。なぜなら、２つのＭＬは、１ヌクレオチド（下線部）相違し、ＭＬ ＧＴＣＡＡＡＴＴは、ＧＴＣＡＡＡＡＴより低いＭＬカウントを有するからである。次に、ＭＬ

は、ＧＴＣＡＡＡＡＴより高いＭＬカウントを有するＭＬ

（ＭＬ配列中の相違を下線で示す）へ折りたたまれうる。同様に、ＭＬ ＴＴＣＡＧＡＡＡおよびＣＴＣＡＡＡＡＡは、ＭＬ ＴＴＣＡＡＡＡＡへ折りたたまれうる。ＭＬ ＴＴＣＡＡＡＣＴは、ＭＬ ＴＴＣＡＡＡＡＴへ折りたたまれ、これが、今度は、ＭＬ ＴＴＣＡＡＡＡＡに折りたたまれうる。ＭＬ ＴＴＣＡＡＡＣＡは、他のすべてのＭＬと２ヌクレオチド以上相違するため、他の８つのＭＬのいずれにも折りたたまれない。ＲＳＥＣ訂正前に、未補正ＭＬカウントは９であった。ＲＳＥＣ訂正後、ＭＬカウントは２つ：ＭＬ ＴＴＣＡＡＡＡＡおよびＴＴＣＡＡＡＡＡであった。 FIG. 31 shows a schematic diagram of an example of recursive replacement error correction outlined above. The ML in the uncorrected data before RSEC correction includes nine unique MLs: GTCAAAATT, GTCAAAAT, GTCAAAAA, TTCAAAAA, TTCAGAAA, CTCAAAAA, TTCAAAACT, TTCAAAAT, and TTCAAACA. By applying RSEC

teeth,

Can be folded into. This is because the two MLs differ by one nucleotide (underlined) and the ML GTCAAATT has a lower ML count than the GTCAAAATT. Next, ML

Has a higher ML count than GTCAAAAT

Can be collapsed (underlined differences in the ML sequence). Similarly, ML TTCAGAAA and CTCAAAAA can be folded into ML TTCAAAAA. The ML TTCAAAACT can be folded into the ML TTCAAAAT, which in turn can be folded into the ML TTCAAAAA. ML TTCAAACA differs from all other MLs by more than 2 nucleotides and therefore does not fold into any of the other 8 MLs. Prior to the RSEC correction, the uncorrected ML count was 9. After RSEC correction, the ML counts were two: ML TTCAAAAA and TTCAAAAA.

Overall, these data demonstrate the process of using RSEC to correct uncorrected ML counts.

Example 12
ML Coverage Calculation This example illustrates ML coverage calculation.

After RSEC, the gene ML counts per well are evaluated to determine their suitability for further corrections. Genes with low ML coverage (<4 reads per ML) bypass the next correction step and are reported in the final ML data table and recorded as "low depth" in the bioinformatics pipeline. For genes with extremely high inputs, such as at least 6557 of the 6651 possible barcodes observed, the barcode diversity makes it difficult to determine the number of molecules and the gene is considered "saturated". Is displayed. For gene MLs that do not meet either of the two decision points, proceed to the next DBEC algorithm and display "pass" in the output log file. In addition, genes with ML above an average of 650 ML per well are recorded as "high input" because> 5% of these MLs are reused based on the Poisson distribution (FIG. 27).

Overall, this example illustrates ML coverage calculations.

Example 13
Distribution-based error correction (DBEC)
This embodiment illustrates distribution-based error correction.

Unlike RSEC, the DBEC algorithm is a method for identifying whether an ML is an error or a true signal, regardless of its ML sequence. RSEC relies on both ML sequences and ML coverage information to correct errors, while DBEC relies primarily solely on ML coverage to correct for non-replacement error correction. As mentioned above, error barcodes generally have a low ML coverage that is different from the true barcode ML coverage; this difference in ML coverage is seen as a different distribution in the ML coverage histogram plot. Can be done (Fig. 27). Given this difference, the DBEC has two negative binomials to statistically distinguish between ML errors (having lower ML coverage) and those of the true signal with higher ML coverage. Apply the term distribution.

Removal of Reused ML for Optimal Distribution Fitting For a given gene, as the detected ML increases, the ML to be reused (ie, the same ML to label two or more mRNAs from the same gene) Is used), which can be estimated from the increase. Using the Poisson distribution (γ _non-unique ), the number of reused MLs in well i (n _{non-unique, i} ) is estimated from the ML reuse rate equation (equation (6)). If the putative reuse ML is greater than 5% of the total ML of a given gene in well i, then this gene in well i is labeled as "high input". For these "high input" data, the maximum ML coverage ML is excluded from the distribution fit (but preserved for later counting steps) in order to obtain a better binomial distribution.
P (X> 1│λ _non-unique ), λ _non-unique = Wavelength of
ML / 6561 formula (6)

Addition of pseudopoints for underexpressed genes When the unique number of MLs is less than 10, it is often difficult to fit the distribution due to the sparseness of the data. To remedy this problem, DBEC adds a 1% signal count pseudopoint used to aid distribution fitting, but still does not affect the data.

Parameter Estimates Two sets of starting values for parameter estimation are estimated to fit the two negative binomial distributions and distinguish errors from the signal ML. The error distribution is assumed to be a negative binomial distribution with mean and 1 scatter.

Error / Signal Probability Estimation Signals and error distributions are assumed as Negative Binomial (μ _signal , size _signal ) and Negative Binomial (μ _error , size _error ), respectively. To determine the number of signal MLs in ascending order, the probability that the number of reads from a given ML will derive from the signal and error distribution is calculated until equation (8) is satisfied, where the preceding ML Are all considered error MLs.
P (X = r│μ = μ _error , size = size _error ) <P (X = r│μ = μ _signal , size = size _signal ) Equation (8)

Overall, this example shows a calculation for performing distribution-based error correction.

Example 14
Adjustment of SL Error Based on Second Derivative Function This example shows adjustment of SL error based on the second derivative.

32 (a)-(e) show exemplary results of PCR and sequencing error correction based on the second derivative of molecular labeling depth changes. FIG. 32, panel (a) shows that SL error and signal ML can be well separated. 32, panels (b) and (d) show the cumulative sum of molecularly labeled counts from the ML counts shown in FIGS. 32, (c) and (e), respectively. In FIG. 32, the vertical lines in the panels (b) and (d) indicate the position of the maximum value of the quadratic derivative. Dotted lines in FIGS. 32b and (d) indicate that the position of the maximum value of the quadratic derivative can separate the ML in the ML count vs. ML read depth plot.

Overall, these data reveal that the maximum value of the quadratic derivative of the molecular label can be used to isolate the SL error from the ML signal.

Example 15
Correction of PCR and Sequencing Errors Based on DBEC This example shows correction of PCR and sequencing errors based on two negative binomial distributions.

33, panels (a)-(c), show exemplary results of PCR and sequencing error correction based on two negative binomial distributions for CD69. FIG. 33, panel (a) shows two negative binomial distributions (D _n of noise negative binomial distribution and signal binomial) for CD69 in the ML count data shown in the histogram of ML depth in FIG. 33, panel (b). The fit of D _s ) of the distribution is shown. The dotted line in FIG. 33, panel (b) shows the separation of ML signals and SL errors as determined by the two negative binomial distributions shown in FIG. 33, panel (a). In FIG. 33, the vertical line in panel (c) shows the local maximum of the quadratic derivative determined based on the cumulative sum plot of leads. Similar to FIG. 33, FIGS. 34, (a)-(c) show exemplary results of PCR and sequencing error correction based on two negative binomial distributions for CD3E.

Overall, these data reveal that DBEC can be used to correct PCR and sequencing errors.

Example 16
ML Reuse This example demonstrates the need to recycle ML for highly expressed genes, as well as to regulate the input data for highly expressed genes prior to distribution fitting.

35, panels (a)-(c) show exemplary results of PCR and sequencing error correction based on two negative binomial distributions for the highly expressed gene ACTB. Highly expressed genes can have over-sequencing status (eg, have an ML coverage of 100 or greater). In some embodiments, the highly expressed gene may be determined using other criteria. In FIG. 35, panel (a), the molecular label to the right of the vertical line corresponds to the ML that was probably reused based on the high depth. FIG. 35, panel (b) schematically shows that molecular labeling can be divided into three categories (other than ML errors): SL errors, signal MLs, and possibly reused MLs. FIG. 35, panel (c) demonstrates that the fitted two negative binomial distributions were not ideal, probably without adjusting for the reused ML.

FIG. 36 shows exemplary results of reuse of G-rich molecular labels for highly expressed genes. FIG. 36 shows the top 20 high depth MLs for the highly expressed genes GAPDH, ACTB, and HSP90AB1. These high depth MLs had a large number of Gs and Ts, which were likely to be reused and the bar coding was not stochastic. The ML doublet occurred earlier than the theoretically calculated value assuming the probability indicator. For ACTB, if there were 350 ML per well, there should have been a 2.7% doublet in theory, but the actual doublet was around 4%.

37, panels (a)-(b) show exemplary results of regulation of input data for highly expressed genes prior to fitting the two negative binomial distributions. 37, panel (a) shows the input data in FIG. 35, panel (a), regulated for highly expressed genes. In contrast to the non-ideal distribution fit in FIG. 35, panel (c), FIG. 37, panel (b) shows the two negative binomial distributions fitted.

Overall, these data indicate that reused ML may need to be removed from sequenced data for highly expressed genes prior to fitting the two negative binomial distributions.

Example 17
Correction of ML Counts Using Two Negative Binomial Distributions This example shows the ML counts of 10 targets corrected using two negative binomial distributions.

38, panels (a)-(j), show a non-limiting exemplary validation of a data set corrected using two negative binomial distributions. As shown in FIG. 38, the ML counts for 10 targets have been corrected. The vertical lines in each panel of FIG. 38 show the separation of the target ML signal and SL error as determined using the two negative binomial distributions.

Overall, these data validate ML count corrections using two negative binomial distributions.

Example 18
T-Probabilistic Proximal Implantable Visualization of BD Precise ™ Targeted Assay from 96 Wells of Mixed Jurkat and Breast Cancer (BrCa) Single Cells This example is a mixed Jurkat and Breast Cancer (BrCa) Single Cell. Demonstrates how to correct PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction for.

39, panels (a)-(d) are exemplary t-probabilistic neighborhood implants of BD Precise ™ Targeted Assay from 96 wells of mixed Jurkat and breast cancer (BrCa) single cells. SNE) Shows visualization (86 test genes). FIG. 39, panel (a) shows that cell clusters were identified using DBScan with the same parameters before and after ML regulation. 39, panels (b)-(d) show individual marker expression assessed by both color and point size. FIG. 39, panel (b) shows PSMB4, a housekeeping gene present in both cell types and after ML regulation, and the lack of PSMB4 signal is further accentuated in the "low signal" cluster. FIG. 39, panel (c) shows CD3E, a lymphocyte marker that emphasizes Jurkat cell clusters. FIG. 39, panel (d), shows CDH1, an epithelial cell marker that emphasizes BrCa clusters.

Overall, these data demonstrate that ML regulation removed ML noise, which allowed a clear distinction of gene expression between cell clusters.

Example 19
Analysis of Difference Expression Between Cell Clusters This example shows how to correct PCR and sequencing errors based on recursive replacement error correction and distribution-based error correction for low signal cells and breast cancer (BrCa) cells.

40, panels (a)-(b) show the difference expression analysis between cell clusters for genes with> 0 ML in both selected clusters, calculated by DBScan in each cluster and determined by the genetic marker level. It is a non-limiting exemplary plot shown. FIG. 40, panel (a), shows "low signal" cluster gene expression compared to the remaining cells. In FIG. 40, the upper part of the panel (a) shows an uncorrected ML comparison, which shows that genes with higher average expression in other cells generally have higher ML noise. FIG. 40, lower part of panel (a) reduces ML noise detected in "low signal" clusters after ML regulation with RSEC and DBEC, allowing clear identification of gene expression between clusters. Show that. FIG. 40, panel (b), shows the development of the "BrCa" cluster gene compared to the remaining cells. FIG. 40, upper part of panel (b) shows that uncorrected MLs in non-BrCa cells also had significant ML counts for BrCa markers such as KRT1, MUC1. In FIG. 40, the lower part of the panel (b) shows that the regulated ML of the BrCa marker was much more abundant in the BrCa cluster than the remaining cells.

Overall, these data show that for cells such as low signal cells and breast cancer cells, PCR and sequencing errors can be corrected based on recursive replacement error correction and distribution-based error correction.

Example 20
Modulation of Molecular Labeling of Mixed Jurkat and T47D Cells This example shows a method of regulating molecular labeling of mixed Jurkat and T47D cells.

41, panels (a)-(d) show t-stochastic near-implantation visuals of BD Precise ™ Targeted Assay from 96 wells of mixed Jurkat and breast cancer (T47D) single cells containing 86 test genes. It is a non-limiting exemplary plot showing the transformation. FIG. 41, panel (a) shows that cell clusters were identified using DBScan with the same parameters before and after ML regulation. 41, panels (b)-(d), show individual marker expression assessed by both color and point size. FIG. 41, panel (b) shows the evaluation of PSMB4, a housekeeping gene present in both cell types and after ML regulation. The lack of PSMB4 signals is further accentuated in untemplated control (NTC) clusters. FIG. 41, panel (c), shows the evaluation of CD3E, a lymphocyte marker that emphasizes Jurkat cell clusters. FIG. 41, panel (d), shows the evaluation of CDH1, an epithelial cell marker that emphasizes T47D clusters.

42, panels (a) to (b) are before the error correction step (uncorrected ML shown in FIG. 42, panel (a)) and after RSEC and DBEC correction (adjustment ML shown in FIG. 42, panel (b)). In addition, it is a non-limiting exemplary heat map showing the difference gene expression by molecular labeling count between the various cell clusters identified in FIG. 41. Genes with low expression are blue and genes with high expression are orange. Genes with similar gene expression among these cell types cluster together. In the absence of error correction, NTC had noise from highly expressed genes such as CD3E and KRT18 (Jurkat and T47D markers, respectively). In addition, error correction revealed a gene expression pattern that could be distinguished between Jurkat and T47D.

Overall, these data demonstrate that ML regulation can eliminate MI noise, which allows for a clear distinction of gene expression between cell clusters.

In at least some of the embodiments described above, the one or more elements used in the embodiment can be used interchangeably with other embodiments. However, only if such replacement is technically feasible. Those skilled in the art will appreciate that various other omissions, additions, and modifications to the methods and structures described above may be made without departing from the scope of the claimed subject matter. All such changes and changes are intended to be within the scope of the subject matter set forth in the appended claims.

As appropriate in the context and / or application in connection with the use of substantially any plural and / or singular term described herein, those skilled in the art may from plural to singular and /. Alternatively, conversion from the singular to the plural is possible. Various singular / plural interchanges may be explicitly described herein for clarity. As used herein and in the appended claims, the singular "a," "an," and "the" include multiple references, unless expressly specified in the context. Will be done. Any meaning of "or (or)" herein is intended to include "and / or (and / or)" unless otherwise stated.

In general, it is generally intended by those skilled in the art that the terms used herein, in particular in the appended claims (eg, the text of the appended claims), are "open" terms. As will be understood (for example, the term "inclusion" should be interpreted as "including but not limited to" and the term "having" should be interpreted as "having". It should be interpreted as "having at least", and the term "includes" should be interpreted as "including but not limited to"). Further, it will be appreciated by those skilled in the art that if a particular number of introductory claim recitations are intended, such intent will be explicitly resited in the claim and that such intent does not exist in the absence of such recitation. For example, to aid understanding, the following claims include the use of the introductory terms "at least one" and "one or more" to introduce claim recitation. Can be included. However, even if such a phrase is used, the introduction of claim recitation by the indefinite article "a" or "an" includes any particular claim comprising such introductory claim recitation, one such recitation. It should not be construed to mean limiting to embodiments only. Even if the same claim contains the introductory phrase "one or more" or "at least one" and an indefinite article such as "a" or "an". (For example, "a" and / or "an" should be interpreted to mean "at least one" or "one or more") .. The same is true when introducing claim recitation using definite articles. Other than that, one of ordinary skill in the art will know that even if a certain number of introductory claim recitations are explicitly resited, such recitation should be construed to mean at least the number of resites (). For example, bear recitation without the other modifier "two recitations" means at least two recitations or two or more recitations). Further, when conditions similar to "at least one of A, B, and C" are used, such configurations are generally intended to mean that one of ordinary skill in the art will understand the conditions (eg, for example. A "system having at least one of A, B, and C" is, but is not limited to, A alone, B alone, C alone, both A and B, both A and C, B and C. Will include systems with both and / or all of A, B and C, etc.). When conditions similar to "at least one of A, B, or C, etc." are used, such configurations are generally intended to mean that one of ordinary skill in the art understands the conditions (eg, "" A system having at least one of A, B, or C "is, but is not limited to, A alone, B alone, C alone, both A and B, both A and C, both B and C. , And / or a system having all of A, B, C, etc.). In addition, virtually any selective term and / or phrase representing two or more alternative terms, whether in the description, claims, or drawings, is one of the terms, any of the terms, or the term. It will be understood by those skilled in the art that it should be understood that the possibility of including both is intended. For example, the phrase "A or B" may be understood to include the possibility of "A" or "B" or "A and B".

In addition, if the features or aspects of the disclosure are described by a Markush group, it will be appreciated by those skilled in the art that the disclosure will be described by any individual member or subgroup of members of the Markush group. Will.

As will be appreciated by those skilled in the art, all scopes disclosed herein include all possible subranges and combinations of subranges thereof for any purpose, eg, with respect to the provision of the specification. do. Any of the listed ranges can be easily recognized as being well described and having at least two equal parts, three equal parts, four equal parts, five equal parts, ten equal parts and the like. .. For example, without limitation, each range considered herein can be easily decomposed into a lower third, a middle third, and an upper third. Similarly, as those skilled in the art will understand, expressions such as "to", "at least", "super", "less than" all include the number of sites that have been resited. As discussed, it means a range that can be subsequently decomposed into sub-ranges. Ultimately, as will be appreciated by those skilled in the art, the scope includes each individual member. Thus, for example, a group with 1 to 3 articles means a group with 1, 2 or 3 articles. Similarly, a group with 1 to 5 articles means a group with 1, 2, 3, 4, or 5 articles, and so on.

Various embodiments and embodiments have been disclosed herein, but other embodiments and embodiments will be obvious to those of skill in the art. The various embodiments and embodiments disclosed herein are for purposes of illustration only and are not intended to be limiting, and the true scope and intent are set forth in the following claims.
The following aspects of the present invention are also preferable.
[1] A method for determining the number of targets.
(A) A step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes has a molecular label. With the process including;
(B) The step of acquiring the sequencing data of the target with the probability barcode;
(C) For one or more of the plurality of targets:
(I) A step of counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data;
(Ii) With the step of identifying the cluster of the molecular label of the target using directional proximity;
(Iii) With the step of collapsing the sequencing data obtained in (b) using the cluster of the molecular label of the target identified in (ii);
(Iv) In the step of estimating the number of the targets, the estimated number of the targets is the said in the sequencing data counted in (i) after folding the sequencing data of (ii). A step that correlates with the number of molecular labels with identifiable sequences associated with the target, and
Including, how.
[2] The method according to [1], wherein the plurality of targets include targets for all transcriptomes of cells.
[3] The method according to any one of [1] to [2], wherein the molecular labels of the targets in the cluster are within a predetermined directional proximity threshold value of each other.
[4] The method according to [3], wherein the directional proximity threshold is 1 Hamming distance.
[5] The molecular label of the target in the cluster includes one or more parent molecule labels and a child molecule label of the one or more parent molecule labels, and the number of occurrences of the parent molecule labels is predetermined. The method according to any one of [1] to [4], which is equal to or greater than the directional proximity occurrence number threshold.
[6] The method according to [5], wherein the predetermined number of occurrences of directional proximity threshold value is 2 × ( number of occurrences of child molecule labels) -1 .
[7] The step of collapsing the sequencing data obtained in (b) using the cluster of the molecular label of the target identified in (ii) is
The method according to any one of [1] to [6], which comprises a step of assigning the number of generations of the child molecule label to the parent molecule label.
[8] The method according to any one of [1] to [7], further comprising a step of determining the sequencing depth of the target.
[9] When the sequencing depth of the target exceeds a predetermined sequencing depth threshold, the step of estimating the number of the targets includes the step of adjusting the sequencing data counted in (i). 8].
[10] The method according to [9], wherein the predetermined sequencing depth threshold is 15 to 20.
[11] The step of adjusting the sequencing data counted in (i) is
[9]-[. 10] The method according to any one of the items.
[12] The method according to [11], wherein the step of thresholding the molecular label of the target includes a step of performing statistical analysis on the molecular label of the target.
[13] The step of carrying out the statistical analysis is
A step of applying the distribution of the molecular labels of the target and the number of their occurrences to the two negative binomial distributions;
The step of determining the number n of true molecular labels using the two negative binomial distributions;
The step of removing the false molecular label from the sequencing data obtained in (b), and
Including
The false molecular label comprises a molecular label having a lower number of occurrences than the nth richest molecular label, and the true molecular label has more than the number of occurrences of the nth richest molecular label. The method according to [12], which comprises a molecular label having.
[14] The negative binomial distribution comprises a first negative binomial distribution corresponding to the true molecular label and a second negative binomial distribution corresponding to the false molecular label, according to [13]. the method of.
[15] A method for determining the number of targets.
(A) A step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes has a molecular label. With the process including;
(B) The step of acquiring the sequencing data of the target with the probability barcode;
(C) For one or more of the plurality of targets:
(I) A step of counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data;
(Ii) The step of determining the number of noise molecule labels having an identifiable sequence associated with the target in the sequencing data;
(Iii) The step of estimating the number of the targets and
Including
The identification associated with the target in the sequencing data counted in (i), where the estimated number of the targets was adjusted according to the number of the noise molecule labels determined in (ii). A method that correlates with the number of molecular labels having possible sequences.
[16] The method of [15], further comprising determining the sequencing status of the target in the sequencing data.
[17] The method of [16], wherein the sequencing status of the target in the sequencing data is saturated sequencing, undersequencing, or oversequencing.
[18] The method of [17], wherein the saturation sequencing status is determined by said target having a number of molecular labels containing identifiable sequences greater than a predetermined saturation threshold.
[19] The method of [18], wherein the predetermined saturation threshold is about 6557 where the probability barcode comprises about 6651 molecular labels having an identifiable sequence.
[20] The item according to any one of [18] to [19], wherein when the probability barcode contains about 65536 molecular labels having an identifiable sequence, the predetermined saturation threshold is about 65532. the method of.
[21] When the sequential status of the target in the sequencing data is the saturated sequencing status, the number of the noise molecule labels determined in (ii) is zero, [17] to [17]. The method according to any one of [20].
[22] The undersequencing status is determined by the target having a depth less than a predetermined undersequencing threshold, and the depth of the subject is an identifiable sequence associated with the target in the sequencing data. The method according to any one of [17] to [21], which comprises the average, minimum, or maximum depth of the molecular label having the above.
[23] The method according to [22], wherein the undersequencing threshold is about 4.
[24] The method of [23], wherein the undersequencing threshold is independent of the number of the molecular labels having an identifiable sequence.
[25] When the sequential status of the target in the sequencing data is the under-sequencing status, the number of the noise molecule labels determined in (ii) is zero, [17] to The method according to any one of [24].
[26] The excess sequencing status is determined by the target having a depth greater than a predetermined excess sequencing threshold, and the depth of the subject is an identifiable sequence associated with the target in the sequencing data. The method according to any one of [17] to [25], which comprises the average, minimum, or maximum depth of the molecular label having the above.
[27] The method of [26], wherein the excess sequencing threshold is about 250 when the probability barcode contains about 6651 molecular labels having an identifiable sequence.
[28] When the sequencing data of the target in the sequencing data is the excessive sequencing status.
Any of [26]-[27], further comprising subsampling the number of the molecular labels having an identifiable sequence associated with the target in the sequencing data to the predetermined excess sequencing threshold. The method described in item 1.
[29] The step of determining the number of the noise molecule labels having an identifiable sequence associated with the target in the sequencing data is:
If the negative binomial distribution fitting condition is met,
(Iv) A step of fitting the signal negative binomial distribution to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), wherein the signal negative. With the step corresponding to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), where the binomial distribution is a signal molecular label;
(V) A step of fitting the noise-negative binomial distribution to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), wherein the noise-negative binomial distribution is applied. With the step corresponding to the number of molecular labels having an identifiable sequence associated with the target in the sequencing data counted in (i), where the binomial distribution is a noise molecular label;
(Vi) The step of determining the number of noise molecule labels using the signal negative binomial distribution fitted in (v) and the noise negative binomial distribution fitted in (vi).
The method according to any one of [17] to [28].
[30] The negative binomial distribution fitting condition comprises the fact that the sequencing status of the target in the sequencing data is not the under-sequencing status or the over-sequencing status. Method.
[31] The step of determining the number of the noise molecule labels using the signal negative binomial distribution fitted in (v) and the noise negative binomial distribution fitted in (vi) is
For each of the identifiable sequences associated with the target in the sequencing data
With the step of determining that the signal probability of the identifiable sequence is the signal negative binomial distribution;
With the step of determining that the noise probability of the identifiable array is the noise negative binomial distribution;
If the signal probability is less than the noise probability, then the step of determining that the identifiable sequence is a noise molecule label.
The method according to any one of [29] to [30], which comprises.
[32] The step of determining the number of noise molecule labels having an identifiable sequence associated with the target in the sequencing data is:
The sequencing status of the target in the sequencing data is not the under-sequencing status or the excess-sequencing status and is associated with the target in the sequencing data counted in (i). If the number of the molecular labels with identifiable sequences is less than the pseudo-point threshold, (ii) determines the number of noise molecular labels with identifiable sequences associated with the target in the sequencing data. A step of adding a pseudo-point to the number of the molecular labels having an identifiable sequence associated with the target in the sequencing data.
The method according to any one of [17] to [31].
[33] The method according to [32], wherein the pseudo-point threshold is 10.
[34] The step of determining the number of the noise molecule labels having an identifiable sequence associated with the target in the sequencing data is:
The sequencing status of the target in the sequencing data is not the under-sequencing status or the over-sequencing status and is associated with the target in the sequencing data counted in (i). If the number of the molecular labels having an identifiable sequence is greater than or equal to the pseudo-point threshold, in (ii) the number of the noise molecular labels having the identifiable sequence associated with the target in the sequencing data. Including the step of removing the non-unique molecular label in the determination,
The method according to any one of [17] to [33].
[35] When the step of removing the non-unique molecular label has a number of the molecular labels having an identifiable sequence associated with the target in the sequencing data greater than a predetermined reused molecular label threshold. In (34), the step of removing the non-unique molecular label is included in determining the number of the noise molecular labels having an identifiable sequence associated with the target in the sequencing data. The method described.
[36] The method of [35], wherein the reusable molecular labeling threshold is about 650 when the probability barcode comprises about 6651 molecular labels having an identifiable sequence.
[37] The step of removing the non-unique molecular label is
With the step of determining the theoretical number of non-unique molecular labels for the number of the molecular labels having an identifiable sequence associated with the target in the sequencing data;
A step of removing a molecular label having a higher incidence than the nth abundant molecular label having an identifiable sequence associated with the target in the sequencing data.
Including
The method according to any one of [34] to [36], wherein n is a theoretical number of non-unique molecular labels.
[38] With a hardware processor
When executed by the hardware processor, a non-transient memory storing an instruction to cause the processor to execute the method according to any one of [1] to [37].
A computer system for determining the number of targets, including.
[39] A computer reading medium comprising a software program containing a code for executing the method according to any one of [1] to [37].
[40] A method for determining the number of targets.
(A) A step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes has a molecular label. With the process including;
(B) The step of acquiring the sequencing data of the target with the probability barcode;
(C) For one or more of the plurality of targets:
(I) A step of counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data;
(Ii) With the step of determining the quality status of the target in the sequencing data obtained in (b);
(Iii) A step of determining one or more sequencing data errors in the sequencing data obtained in (b), wherein the one or more sequencing data errors in the sequencing data are determined. The steps are as follows: the number of the molecular labels having an identifiable sequence associated with the target in the sequencing data, the quality status of the target in the sequencing data, and the plurality of probability barcodes. A step comprising determining one or more of the number of said molecular labels having an identifiable sequence of;
(Iv) In the step of estimating the number of the targets, the estimated number of the targets is adjusted according to the one or more sequencing data errors determined in (iii), (i). A step that correlates with the number of the molecular labels having an identifiable sequence associated with the target in the sequencing data counted in.
Including, how.
[41] The method according to [40], further comprising a step of collapsing the sequencing data obtained in (b) before determining the one or more sequencing data errors.
[42] The step of folding the sequencing data obtained in (b) is
2 . One copy has a similar molecular label if the sequences of the molecular labels on the two copies of the target differ by at least one base.
The method according to [41].
[43] The method according to [42], wherein the predetermined folding occurrence number threshold is 7 when the probability barcode contains about 6651 molecular labels having an identifiable sequence.
[44] The method according to [42], wherein the predetermined folding occurrence number threshold is 17 when the probability barcode contains about 65536 molecular labels having an identifiable sequence.
[45] Any one of [42] to [44], wherein the two copies of the target have similar molecular labels when the sequences of the molecular labels of the two copies of the target differ by at least one base. The method described in.
[46] The method according to any one of [40] to [45], wherein the molecular label contains 5 to 20 nucleotides.
[47] The method according to any one of [40] to [46], wherein the molecular labels having different probability barcodes are different from each other.
[48] The method according to any one of [40] to [47], wherein the plurality of probability barcodes contain about 6651 molecular labels having an identifiable sequence.
[49] The method according to any one of [40] to [47], wherein the plurality of probability barcodes include a molecular label of about 65536 having an identifiable sequence.
[50] The method according to any one of [40] to [49], wherein the sequencing data comprises a sequence of the plurality of targets having a read length of 50 nucleotides or more.
[51] The method according to any one of [40] to [49], wherein the sequencing data comprises a sequence of the plurality of targets having a read length of 75 nucleotides or more.
[52] The method according to any one of [40] to [49], wherein the sequencing data comprises a sequence of the plurality of targets having a read length of 100 nucleotides or more.
[53] The sequencing data obtained in (b) can be generated by performing polymerase chain reaction (PCR) amplification on the plurality of probabilistic barcoded targets, [40]-[. 52] The method according to any one of paragraphs.
[54] The one or more sequencing data errors are PCR introduction errors, sequencing introduction errors, errors due to bar code contamination, library production errors, or any combination thereof. [40]-[ 53] The method according to any one of paragraphs.
[55] The method of [54], wherein the PCR introduction error is the result of a PCR amplification error, a PCR amplification bias, inadequate PCR amplification, or any combination thereof.
[56] The method of any one of [54]-[55], wherein the sequencing introduction error is the result of inaccurate base calling, inadequate sequencing, or any combination thereof.
[57] The method according to any one of [40] to [56], wherein steps (i), (ii), (iii), and (iv) are carried out for each of the plurality of targets.
[58] The method according to any one of [40] to [57], wherein the quality status of the target in the sequencing data is complete sequencing, incomplete sequencing, or saturated sequencing.
[59] The quality status of the target in the sequencing data is associated with the number of the molecular labels having identifiable sequences in the plurality of probability barcodes and the target in the counted sequencing data. 58. The method according to [58], which is determined by the number of said molecular labels having an identifiable sequence.
[60] The complete sequencing quality status is determined by a dispersal index compared to the Poisson distribution above a predetermined complete sequencing dispersal threshold, and the predetermined complete sequencing dispersal threshold is 0.9 [58]. ] To [59]. The method according to any one of the following items.
[61] The method according to [60], wherein the predetermined complete sequencing spray threshold is 1.
[62] The method according to [60], wherein the predetermined complete sequencing application threshold is 4.
[63] The complete sequencing quality status is further determined by a molecular label having an occurrence number equal to or greater than a predetermined complete sequencing occurrence number threshold in the sequencing data obtained in (b), and the predetermined complete sequence is further determined. The method according to any one of [60] to [62], wherein the threshold for the number of occurrences of singing is 10.
[64] The method according to [63], wherein the predetermined complete sequencing occurrence number threshold value is 18.
[65] Any one of [58]-[64], wherein the saturation sequencing quality status is determined by the target having a number of molecular labels containing an identifiable sequence greater than a predetermined saturation threshold. The method described in.
[66] The saturation sequencing quality status is further determined by one other target of the plurality of targets having a number of molecular labels containing an identifiable sequence greater than the predetermined saturation threshold. , [65].
[67] The method according to [65], wherein the predetermined saturation threshold is 6557 when the probability barcode comprises about 6651 molecular labels having an identifiable sequence.
[68] The method according to [65], wherein the predetermined saturation threshold is 65532 when the probability barcode contains about 65536 molecular labels having an identifiable sequence.
[69] The quality status of the target in the sequencing data is such that the quality status of the target in the sequencing data obtained in (b) is not complete sequencing and saturated sequencing. The method according to any one of [40] to [68], which is classified as incomplete sequencing if not present.
[70] In (iv), the number of the molecular labels having the identifiable sequence associated with the target in the sequencing data counted in (i).
If the target has the complete sequencing quality status
With the step of determining all child molecule labels for one or more parent molecule labels;
With the step of performing a first statistical analysis on at least one child molecule label and said parent molecule label;
If the null hypothesis of the first statistical analysis is acceptable, the step of assigning the number of occurrences of the child molecule label to the parent molecule label and
The method according to any one of [50] to [69], which is adjusted by.
[71] The one or more parent molecule labels include a molecular label having an occurrence number equal to or higher than a predetermined complete sequencing parent threshold value, and the predetermined complete sequencing parent threshold value is the predetermined complete sequencing occurrence number threshold value. The method according to [70], which is equal to.
[72] The child molecule label comprises a molecule label that is one base different from the parent molecule label and has a number of occurrences that is less than or equal to the predetermined complete sequencing child threshold, and the predetermined complete sequencing child threshold is 3. The method according to any one of [70] to [71].
[73] The method according to [72], wherein the predetermined complete sequencing child threshold is 5.
[74] When the probability that the null hypothesis is true is less than the false discovery rate, the null hypothesis of the first statistical analysis is accepted and the false discovery rate is 5%, [70] to The method according to any one of [73].
[75] The method according to [74], wherein the false discovery rate is 10%.
[76] The method according to any one of [70] to [75], wherein the first statistical analysis is a multiplex binomial test.
[77] In (iv), the number of the molecular labels having the identifiable sequence associated with the target in the sequencing data counted in (i) is.
If the target has the complete sequencing quality status
Adjusted by the step of thresholding the molecular label of the target to determine the true and false molecular label associated with the target in the sequencing data obtained in (b). [50] The method according to any one of [76].
[78] The method according to [77], wherein the step of thresholding the molecular label of the target includes a step of performing a second statistical analysis on the molecular label of the target.
[79] The step of carrying out the second statistical analysis is
A step of applying the distribution of the molecular labels of the target and the number of their occurrences to the two Poisson distributions;
The step of determining the number n of true molecular labels using the two Poisson distributions;
The step of removing the false molecular label from the sequencing data obtained in (b), and
Including
The false molecular label comprises a molecular label having a lower number of occurrences than the nth richest molecular label, and the true molecular label is greater than or equal to the number of occurrences of the nth richest molecular label. The method according to [78], comprising a molecular label having a number of occurrences .
[80] The method according to [79], wherein the two Poisson distributions include a first Poisson distribution corresponding to the true molecular label and a second Poisson distribution corresponding to the false molecular label.
[81] In (iv), the number of the molecular labels having the identifiable sequence associated with the target in the sequencing data counted in (i).
When the quality status of the target in the sequencing data obtained in (b) is the incomplete sequencing quality status.
The step of determining whether or not the target is noisy in the sequencing data obtained in (b);
The step of removing the noisy target from the sequencing data obtained in (b), and
The method according to any one of [58] to [80], which is adjusted by.
[82] If the number of occurrences of the molecular label of the noisy target is equal to or less than the incomplete sequencing noisy target threshold value, the target is noisy and the incomplete sequencing noisy gene threshold value is 5. The method according to [81].
[83] The method of [82], wherein the incomplete sequencing noisy target threshold is equal to the central number of occurrences of the molecular label of the plurality of targets having a quality status of complete sequencing.
[84] The method according to [82], wherein the incomplete sequencing noisy target threshold is equal to the average number of occurrences of the molecular label of the plurality of targets having a quality status of complete sequencing.
[85] In (iv), the number of the molecular labels having the identifiable sequence associated with the target in the sequencing data counted in (i).
When the quality status of the target in the sequencing data obtained in (b) is the incomplete sequencing quality status.
[50]-[84], wherein the molecular label of the target is thresholded to determine a true molecular label and a false molecular label in the sequencing data obtained in (b). The method described in any one of the items.
[86] The method according to [85], wherein the step of thresholding the molecular label of the target comprises a step of performing a third statistical analysis on the molecular label.
[87] The step of carrying out the third statistical analysis of the molecular label is
With the step of determining the number n of true molecular labels using a zero-cut Poisson model;
The step of removing the false molecular label from the sequencing data obtained in (b), and
Including
The false molecular label comprises a molecular label having a lower number of occurrences than the nth richest molecular label, and the true molecular label has more than the number of occurrences of the nth richest molecular label. The method according to [86], comprising a molecular label having a number .
[88] The sequencing obtained in (b) after the sequencing data counted in (i) is adjusted according to the one or more sequencing data errors determined in (iii). The method according to any one of [40] to [87], wherein at least 50% of the molecular label in the data is retained.
[89] The sequencing data counted in (i) is adjusted according to the one or more sequencing data errors determined in (iii), and then obtained in (b) b). The method according to any one of [40] to [87], wherein at least 80% of the molecular label in the sequencing data is retained.
[90] The step of attaching a probability barcode to the plurality of targets includes a step of hybridizing the plurality of probability barcodes with the plurality of targets to generate the target with the probability barcode [40]. The method according to any one of [87].
[91] The method according to [89], wherein the step of attaching a probability barcode to the plurality of targets includes a step of creating an indexed library of the target with the probability barcode.
[92] In any one of [89] to [91], the step of producing the indexed library of the target with the probability barcode is carried out using the solid carrier containing the plurality of probability barcodes. The method described.
[93] The method according to [92], wherein the solid carrier comprises a plurality of synthetic particles bound to the plurality of probability barcodes.
[94] Each of the plurality of probability barcodes comprises one or more of a sample label, a universal label and a cell label, wherein the sample label is the same as for the plurality of probability barcodes on the solid carrier. , Universal labeling is the same for the plurality of probability barcodes on the solid carrier, and cell labeling is the same for the plurality of probability barcodes on the solid carrier, [92]-[. 93] The method according to any one of paragraphs.
[95] The method according to [94], wherein the sample label comprises 5 to 20 nucleotides.
[96] The method according to any one of [94] to [95], wherein the universal label comprises 5 to 20 nucleotides.
[97] The method according to any one of [94] to [96], wherein the cell label comprises 5 to 20 nucleotides.
[98] The method according to any one of [92] to [95], wherein the solid carrier comprises the two-dimensional or three-dimensional probability barcodes.
[99] The method according to any one of [93] to [98], wherein the synthetic particles are beads.
[100] The method according to [99], wherein the beads are silica gel beads, regulated porous glass beads, magnetic beads, dyna beads, Sephadex / Sephadex beads, cellulose beads, polystyrene beads, or any combination thereof. ..
[101] The method according to [40] to [100], wherein the solid carrier comprises a polymer, a matrix, a hydrogel, a needle array device, an antibody, or any combination thereof.
[102] The method according to any one of [40] to [101], wherein the plurality of targets are contained in the sample.
[103] The method of [102], wherein the sample comprises one or more cells.
[104] The method according to [102], wherein the sample is a single cell.
[105] The method according to [102], further comprising the step of lysing the one or more cells.
[106] The step of lysing the one or more cells comprises heating the sample, contacting the sample with a detergent, changing the pH of the sample, or any combination thereof [105]. ] The method described in.
[107] The method according to [102], wherein the one or more cells contain one or more cell types.
[108] At least one of the one or more cell types is a brain cell, a heart cell, a cancer cell, a circulating tumor cell, an organ cell, an epithelial cell, a metastatic cell, a benign cell, a primary cell, a circulating cell, or any of them. The method according to [107], which is a combination of the above.
[109] The plurality of targets are ribonucleic acid (RNA), messenger RNA (mRNA), microRNA, small interfering RNA (siRNA), RNA degradation products, RNA containing poly (A) tail, or any of them. The method according to any one of [40] to [108], which comprises a combination.
[110] The method according to any one of [40] to [109], wherein the method is multiplexed.
[111] A method of determining the number of targets.
(A) A step of attaching a probability barcode to a plurality of targets using a plurality of probability barcodes to generate a target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes has a molecular label. With the process including;
(B) The step of acquiring the sequencing data of the target with the probability barcode;
(C) For one or more of the plurality of targets:
(I) A step of counting the number of molecular labels having an identifiable sequence associated with the target in the sequencing data;
(Ii) With the step of identifying the cluster of the molecular label of the target using directional proximity;
(Iii) With the step of collapsing the sequencing data obtained in (b) using the cluster of the molecular label of the target identified in (ii);
(Iv) In the step of estimating the number of the targets, the estimated number of the targets is the said in the sequencing data counted in (i) after folding the sequencing data of (ii). A step that correlates with the number of molecular labels with identifiable sequences associated with the target, and
Including, how.
[112] The method of [111], wherein the plurality of targets include targets for the entire transcriptome of the cell.
[113] The method according to any one of [111] to [112], wherein the molecular labels of the targets in the cluster are within a predetermined directional proximity threshold of each other.
[114] The method according to [113], wherein the directional proximity threshold is a Hamming distance of 1.
[115] The molecular label of the target in the cluster includes one or more parent molecule labels and a child molecule label of the one or more parent molecule labels, and the number of occurrences of the parent molecule labels is predetermined. The method according to any one of [112] to [114], which is equal to or greater than the directional proximity occurrence number threshold.
[116] The method according to [115], wherein the predetermined number of occurrences of directional proximity threshold value is 2 × ( number of occurrences of child molecule labels) -1 .
[117] The step of collapsing the sequencing data obtained in (b) using the cluster of the molecular label of the target identified in (ii) is
The method according to any one of [111] to [116], which comprises a step of assigning the number of occurrences of the child molecule label to the parent molecule label.
[118] The method according to any one of [111] to [117], further comprising a step of determining the sequencing depth of the target.
[119] When the sequencing depth of the target exceeds a predetermined sequencing depth threshold, the step of estimating the number of the targets includes the step of adjusting the sequencing data counted in (i). 118].
[120] The method according to [119], wherein the predetermined sequencing depth threshold is 15 to 20.
[121] The step of adjusting the sequencing data counted in (i) is
[119] to [119], comprising the step of thresholding the molecular label of the target to determine the true molecular label and the false molecular label associated with the target in the sequencing data obtained in (b). 120] The method according to any one of paragraphs.
[122] The method according to [121], wherein the step of thresholding the molecular label of the target includes a step of performing statistical analysis on the molecular label of the target.
[123] The step of carrying out the statistical analysis is
A step of applying the distribution of the molecular labels of the target and the number of their occurrences to the two Poisson distributions;
The step of determining the number n of true molecular labels using the two Poisson distributions;
The step of removing the false molecular label from the sequencing data obtained in (b), and
Including
The false molecular label comprises a molecular label having a lower number of occurrences than the nth richest molecular label, and the true molecular label is greater than or equal to the number of occurrences of the nth richest molecular label. The method according to [122], comprising a molecular label having a number of occurrences .
[124] The method according to [123], wherein the two Poisson distributions include a first Poisson distribution corresponding to the true molecular label and a second Poisson distribution corresponding to the false molecular label.
[125] With a hardware processor
When executed by the hardware processor, the number of targets, including a non-transient memory that stores an instruction to cause the processor to perform the method according to any one of [40] to [124]. Computer system for deciding.
[126] A computer reading medium comprising a software program containing the code for performing the method according to any one of [40] to [124].

Claims (28)

A method of determining the number of nucleic acid targets in a sample ,
(A) A step of attaching a probability barcode to a plurality of nucleic acid targets using a plurality of probability barcodes to generate a nucleic acid target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes is molecularly labeled. including;
(B) A step of acquiring sequencing data of the nucleic acid target with a probability barcode; and (c) the following steps (i) to (iv) for one or more of the plurality of nucleic acid targets:
(I) Counting the number of molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data;
(Ii) A step of identifying a cluster of molecular labels for the nucleic acid target using directional proximity ,
Here, the identifying step comprises retrospectively determining for all molecular labels having an identifiable sequence whether the child molecular label belongs to a cluster containing one or more parent molecular labels, said cluster. The molecular label of the nucleic acid target within contains one or more parent molecule labels and a child molecule label of the one or more parent molecule labels, and the number of occurrences of the parent molecule labels is in a predetermined direction. Proximity occurrence number threshold or higher ;
(Iii) Using the cluster of molecular labels for the nucleic acid target identified in (ii), the step of collapsing the sequencing data obtained in (b); and (iv) estimating the number of the nucleic acid target. Step, the number of said nucleic acid targets estimated herein is identifiable associated with said nucleic acid target in the sequenced data counted in ( i ) after folding the sequenced data in (iii). Correlates with the number of molecular labels with different sequences,
Including, how. The method of claim 1, wherein the molecular labels of the nucleic acid targets in the cluster are within a predetermined directional proximity threshold of each other. The method of claim 2, wherein the directional proximity threshold is a Hamming distance of 1. The method according to claim 1 , wherein the predetermined directional proximity occurrence number threshold value is 2 × ( number of occurrences of child molecule labeling) -1 . The step of collapsing the sequencing data obtained in (b) using the cluster of the molecular label of the nucleic acid target identified in (ii) is
The method according to any one of claims 1 to 4 , which comprises a step of assigning the generation number of the child molecule label to the parent molecule label. The method according to any one of claims 1 to 5 , further comprising a step of determining the sequencing depth of the nucleic acid target. Claimed, wherein when the sequencing depth of the nucleic acid target exceeds a predetermined sequencing depth threshold, the step of estimating the number of the nucleic acid targets comprises adjusting the sequencing data counted in (i). The method according to 6 . The step of adjusting the sequencing data counted in (i) is
7. A step of thresholding the molecular label of the nucleic acid target to determine a true molecular label and a false molecular label associated with the nucleic acid target in the sequencing data obtained in (b). The method described in. The method of claim 8 , wherein the step of thresholding the molecular label of the nucleic acid target comprises performing a statistical analysis of the molecular label of the nucleic acid target. The process of performing the statistical analysis is
A step of applying the distribution of the molecular labels of the nucleic acid target and the number of their occurrences to the two negative binomial distributions;
The step of determining the number n of true molecular labels using the two negative binomial distributions; and the step of removing the sham molecular label from the sequencing data obtained in (b), where the sham The molecular label comprises a molecule label having a lower number of occurrences than the number of occurrences of the nth abundant molecular label, and the true molecular label has a number of occurrences equal to or higher than the number of occurrences of the nth abundant molecular label. 9. The method of claim 9 , comprising a sign. The method of claim 10 , wherein the negative binomial distribution comprises a first negative binomial distribution corresponding to the true molecular label and a second negative binomial distribution corresponding to the false molecular label. A method of determining the number of nucleic acid targets,
(A) A step of attaching a probability barcode to a plurality of nucleic acid targets using a plurality of probability barcodes to generate a nucleic acid target with a plurality of probability barcodes, wherein each of the plurality of probability barcodes is molecularly labeled. including;
(B) A step of acquiring sequencing data of the nucleic acid target with a probability barcode;
(C) A step of determining the sequencing status of the nucleic acid target in the sequencing data;
( D ) For one or more of the plurality of nucleic acid targets, the following (i) to (iii):
(I) Counting the number of molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data;
(Ii) The step of determining the number of noise molecule labels having an identifiable sequence associated with the nucleic acid target in the sequencing data , where the step is.
If the negative binomial distribution fitting condition is met based on the sequencing status
(1) Applying the signal negative binomial distribution to the number of molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data counted in (i), where the signal negative The binomial distribution corresponds to the number of signal molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data counted in (i);
(2) Applying the noise-negative binomial distribution to the number of molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data counted in (i), where the noise-negative The binomial distribution corresponds to the number of molecular labels with identifiable sequences associated with the nucleic acid target in the sequencing data counted in (i), which are noise molecular labels;
(3) To determine the number of the noise molecule labels using the signal negative binomial distribution fitted in (1) and the noise negative binomial distribution fitted in (2).
And (iii ) the step of estimating the number of the nucleic acid targets, wherein the number of the nucleic acid targets estimated herein was adjusted according to the number of the noise molecule labels determined in (iii). Corresponds to the number of molecular labels having the identifiable sequence associated with the nucleic acid target in the sequencing data counted in i).
Including, how. 12. The method of claim 12 , wherein the sequencing status of the nucleic acid target in the sequencing data is saturated sequencing, undersequencing, or oversequencing. 13. The method of claim 13 , wherein the saturation sequencing status is determined by the nucleic acid target having a number of molecular labels comprising an identifiable sequence that is greater than a predetermined saturation threshold. 13 . The method described. The undersequencing status is determined by the nucleic acid target having a depth less than a predetermined undersequencing threshold, and the depth of the nucleic acid target is an identifiable sequence associated with the nucleic acid target in the sequencing data. The method according to any one of claims 13 to 15 , comprising the average, minimum, or maximum depth of the molecular label having. 16. The method of claim 16 , wherein the undersequencing threshold is independent of the number of said molecular labels having an identifiable sequence. 13 to 17 , the number of the noise molecule labels determined in (ii) is zero when the sequencing status of the nucleic acid target in the sequencing data is the undersequencing status. The method described in any one of the items. The excess sequencing status is determined by the nucleic acid target having a depth greater than a predetermined excess sequencing threshold, and the depth of the nucleic acid target is an identifiable sequence associated with the nucleic acid target in the sequencing data. The method according to any one of claims 13 to 18 , comprising the average, minimum, or maximum depth of the molecular label having. When the sequencing data of the nucleic acid target in the sequencing data is the excess sequencing status.
19. The method of claim 19 , further comprising subsampling the number of molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data near the predetermined excess sequencing threshold. .. 12. The method of claim 12 , wherein the negative binomial distribution fitting condition comprises that the sequencing status of the nucleic acid target in the sequencing data is not the under-sequencing status or the over-sequencing status. The step of determining the number of the noise molecule labels using the signal negative binomial distribution fitted in (v) and the noise negative binomial distribution fitted in (vi) is:
For each of the identifiable sequences associated with the nucleic acid target in the sequencing data
With the step of determining that the signal probability of the identifiable sequence is within the signal negative binomial distribution;
With the step of determining that the noise probability of the identifiable array is within the noise negative binomial distribution;
If the signal probability is less than the noise probability, then the step of determining that the identifiable sequence is a noise molecule label.
The method according to any one of claims 12 to 21 , which comprises. The step of determining the number of the noise molecule labels having an identifiable sequence associated with the nucleic acid target in the sequencing data is:
The sequencing status of the nucleic acid target in the sequencing data is not the under-sequencing status or the excess-sequencing status and is associated with the nucleic acid target in the sequencing data counted in (i). If the number of the molecular labels having the identified identifiable sequence is less than the pseudo-point threshold, then in (ii) the noise molecular labels having the identifiable sequence associated with the nucleic acid target in the sequencing data. Prior to determining the number, comprising adding a pseudo-point to the number of the molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data.
The method according to any one of claims 13 to 22 . The step of determining the number of the noise molecule labels having an identifiable sequence associated with the nucleic acid target in the sequencing data is:
The sequencing status of the nucleic acid target in the sequencing data is not the under-sequencing status or the excess sequencing status and is associated with the nucleic acid target in the sequencing data counted in (i). When the number of the molecular labels having the identified identifiable sequences is greater than or equal to the pseudo-point threshold, the noise molecular labels having the identifiable sequence associated with the nucleic acid target in the sequencing data in (ii). Including the step of removing non-unique molecular labels in determining the number of
The method according to any one of claims 13 to 23 . If the step of removing the non-unique molecular label has a number of the molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data greater than a predetermined reused molecular labeling threshold (ii). 24. The step of removing the non-unique molecular label in determining the number of the noise molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data. the method of. The step of removing the non-unique molecular label is
The step of determining the theoretical number of non-unique molecular labels for the number of the molecular labels having an identifiable sequence associated with the nucleic acid target in the sequencing data; and to the nucleic acid target in the sequencing data. The step of removing a molecular label having a larger number of occurrences than the nth most abundant molecular label among the molecular labels having an associated identifiable sequence, where n is the theoretical number of non-unique molecular labels. 24. The method of claim 24 or 25 , including. With a hardware processor
When executed by the hardware processor, a non-transient memory storing an instruction to cause the processor to execute the method according to any one of claims 1 to 26 .
A computer system for determining the number of nucleic acid targets, including. A computer reading medium comprising a software program comprising the code for performing the method according to any one of claims 1 to 26 .

Images